by Dr. Mark White
White & Associates, Laredo, TX (www.oocities.org/wallstreet/7891)
NNCP, UNAM, Mexico City (www.nuclecu.unam.mx/~nncp)
Adaptive Technologies, Mexico City

Abstract: Econophysicists can advance scientific understanding of finance more by developing their own financial research agenda than by following the established Neoclassical agenda. This paper discusses certain empirical regularities that offer a basis for an interesting alternative to the increasingly stylized and sterile debate over such traditional issues as market efficiency. These pervasive phenomena in trader behavior and market design could help econophysicists calibrate agent-based models to breed superior new traders for existing market designs, and test new market designs that can mitigate the excess volatility stemming from existing designs. Trading behavior research should be self-funding with a modicum of capital; market design research might attract funding from market regulators.

A Research Agenda for Agent-Based Finance Models

by Dr. Mark White

Econophysicists have recently begun to coalesce around a serious research program in finance, applying their substantial tool kit to a fresh, potentially fertile terrain (Farmer, 1999). Prominent in this program are efforts to determine empirical statistical regularities in price series, efforts to develop random process models for prices and agent-based models for markets that generate those regularities, and efforts to apply those regularities and models to securities pricing and trading. These current efforts have recognizable roots in the Sixties and Seventies (Mandelbrot, 1963; Schelling 1971; Osborne, 1965), but despite such notable achievements as establishing and modelling clustered volatility and fat tails, their interactions with the mainstream Neoclassical financial research agenda have yet to produce an academic financial science with a significant capacity to predict or transform outcomes. This blame for this shortfall lies less with the econophysicists’ tool kit than with their attempts to climb the existing ivory towers of finance. By responding to the pervasive phenomena that characterize real world finance, rather than the Neoclassical agenda that dominates the financial debate, econophysicists can develop a research agenda that makes real progress in finance.

This paper looks broadly at the finance agenda, and identifies some areas of intense concentration that distract finance academics and econophysicists alike from predicting or transforming outcomes, and other areas of relative neglect with significant potential for scientific productivity. The areas intensely studied tend to respond to issues Neoclassical theory raises (Mirowski, 1989), such as market efficiency, investor rationality, and capital asset pricing models, while the areas relatively neglected tend to respond to primarily empirical agendas, such as the group trading behavior studied by sociologists and journalists (Abolafia, 1984; Baker, 1984; Fishman, 1995), individual trading behavior studied by psychogists (Andreassen, 1988), the manipulative and deviant behavior studied in the legal/regulatory agenda (Friend and Herman, 1965; Lejeune, 1984; Wolfson, 1980) and the market design issues raised by the experimental economics and market technology agendas (Smith, Suchanek, and Williams, 1988; Rietz, 1998; Domowitz and Steil, 1999). These agendas openly address how traders actually behave, rather than looking for specific confirmations of or departures from expected utility axioms, and openly address how market designs and technologies influence that actual trading behavior for better or worse, rather than looking for specific confirmations of or departures from their expected effects on rational traders. With open empirical agendas, these research streams provide econophysicists with far more insight into the pervasive phenomena that their models must account for than the highly derivative, stylized results that stem from the Neoclassical agenda.

Neoclassical theory, as Mirowski documents, has a scientistic or scholastic agenda that focuses on studying equilibrium models derived from 19th century physics, rather than economic reality derived from observation. This Neoclassical agenda has dominated US economics departments for decades, and with its endlessly flexible empirical agenda, could do so for decades more. Whenever an outcome varies from equilibrium predictions, orthodox Neoclassicals simply adds a degree of freedom to account for the variation. Orthodox Neoclassicals make no effort to economize on degrees of freedom, since unlike their theory, all their "epicycles" come from direct observation of economies and hence seemingly add to the plausibility of their models. (It would never occur to them to ask what the theory is for when all the action comes from the bits and pieces they tack on. The answer is too obvious: studying the theory is what makes them economists.)

A good example of adding degrees of freedom in finance comes from closed-end funds, where orthodox Neoclassicals epicyclically "explain" anomalous discounts (relative to the Neoclassical value additivity principle) by by appealing to management fees, or embedded tax liabilities, or any of a number of other superficially plausible factors (Malkiel, 1977). They "explain" anomalous premiums with superior management services, or preferential access to restricted markets, or any of a number of different but superficially plausible factors. They "explain" varying discounts and premiums, when not totally ignoring them, with "random shocks" around the truly relevant equilibrium variable, the mean value of the discount. These random variations occur within arbitrage limits expanded by transactions costs, short-sale restrictions, foreign investment restrictions, and so forth. All these factors indisputably exist, and given their existence, the true Neoclassical can and will believe any constellation of their imputed effects necessary to rescue the value additivity principle from the extreme violence that closed-end fund bubbles and crashes do to it (DeLong and Shleifer, 1991; Rosser, Koppl, Ahmed, and White, 1997), much less the normal violence from the abnormal returns that investors can earn by overweighting large discounts and underweighting premiums in their portfolios (Thompson, 1978).

Given Neoclassical dominance in America’s academe, even those who disagree with its models and assumptions still tend to frame their disagreements in terms of its agenda. Soon after Von Neuman and Morgenstern (1944) advanced the axioms of Neoclassical expected utility theory, Allais began running experiments showing that subjects violated expected utility (Allais, 1953). For generations since, apostate Neoclassical economists and psychologists have run endless variations on such psychological experiments and identified numerous real-world anomalies apparently consistent with their findings (sometimes reporting the same anomalies a generation apart, as in Zweig’s (1973) and Lee, Shleifer, Thaler’s (1991) approach to closed-end funds, or Niederhoffer and Osborne’s (1966) and Cutler, Poterba, and Summer’s (1998) approach to news events, where investors overreact to some news, underreact to other news, and occasionally even react properly to some bit of news). While many individual economists and psychologists have advanced professionally and personally in this prolonged thrust and parry of behavioral finance, within the academic finance profession it has long looked much more like a stylized ritual than a scientfic debate -- whose degrees of freedom shall we add? When the weight of apostate Neoclassical evidence forces orthodox Neoclassical researchers to admit such a fatal exception to investor rationality as framing effects (Tversky and Kahneman, 1986), one that seems to require total abandonment of standard Neoclassical behavioral models, the orthodox profession seems unapologetically ready to retain those models by simply adding another degree of freedom to encompass a framing variable (Machina, 1987).

To continue the closed-end fund example, apostate neoclassicals don’t derive a totally new valuation model from their behavioral observations. They "explain" discounts and premiums by adding an irrational expectations term to standard valuation equations, and show that this term varies inversely through time in ways empirically related to the future market performance of not only closed-end funds themselves (i.e., premia indicate lower future returns, discounts indicate higher future returns; Thompson, 1978), but other assets such as small-capitalization stocks (Zweig, 1973; Lee, Sheifer, and Thaler, 1991). Responding to the apostates, orthodox researchers find one minor point or another subject to methodological criticism (Chen, Kan, and Miller, 1993), and ignore the main point that closed-end fund behavior stands as glaring anomaly to their joint hypothesis of efficient markets and investor rationality. Talking past each other, the orthodox and apostate leave the way open to endless debate about methodological details relating to whose degrees of freedom, letting any prospect of real understanding of the anomalous market behavior of closed-end fund shares recede to the indefinite future.

Econophysics, on the other hand, addresses the closed-end fund issue quite elegantly with its approximations to financial market dynamics stemming from a Neoclassically-agnostic approach to evolving ecologies of trading strategies. Agent-based models following Arthur, Holland, LeBaron, Palmer, and Tayler (1997) and analytic models following Farmer (2000) simply and elegantly capture such well-known and relatively-uncontroversial pervasive phenomena in financial markets as fat tails, and such well-known and relatively-controversial pervasive phenomena as excess volatility. Farmer’s analytic model also shows how prices can depart persistently from fundamental values,. His numerical simlutions of state-dependent threshold value strategies show how real observable strategies (i.e. strategies publicly recommended by respected market analysts) can maintain ongoing discounts to value. Econophysics research methods already model the real financial world, but econophysicists could easily get distracted from real world issues if they let orthodox and apostate Neoclassicals direct their attention to theoretical suppositions rather than observable phenomena.

For example, agent-based and analytic models of evolving trading strategy ecologies can already model trading in single securities accurately, including fat-tails, excess volatility, and long-term departures of trading prices from fundamental values. If econophysicists generally were aware of the robust evidence that traders on trading floors physically flock to markets with high trading volume and volatility, they’d be likely to add multiple securities to their models to gain insights on this important process and the role it may have played in stimulating the tremendous growth in trading volume and volatility over the last few decades. Unfortunately, they’re unlikely to learn about this kind of pervasive phenomena from the mainstream financial economics literature, which is more likely to focus on derivations of the utility-maximizing behavior of a representative agent than observations of flocking behavior among diverse agents. The convoluted empirical debate in financial economics practically forces econophysicists to seek outside the field for the empirical facts to drive their research agenda.

The challenge to econophysicists, then, is to learn more about trading ecologies and their evolution, including cross-market dynamics consistent with empirical observations of trading behavior, and then to integrate that knowledge into their agenda for modelling financial agents and markets. This paper contributes by discussing a few important empirical observations about trading behavior and market design, including evidence for superior traders that suggests that better financial agent models should lead to better price predictions and trading strategies, letting econophysicists earn significant excess returns from their portfolios. Impressive as that feat might be compared to current and foreseeable results of the inbred debate over the Neoclassical agenda, the real test of econophysics will be in developing the knowledge needed to guide the ongoing transformation of existing financial exchanges into new exchanges whose market institutions minimize the excess volatility and mispricing that stems from interactions among trading strategies. Current levels of excess volatility are not an inevitable characteristic of all possible financial markets, but are rather a side effect of the market makers’ pursuit of volume (and hence income) maximization as a positive goal. Econophysics should discover market designs which provide more effective liquidity for most traders while keeping trading volume down to levels consistent with real liquidity needs.

Fundamental values form a backdrop to financial markets, but they are far from the leading force in the volatility we observe. Comparisons of volatility with and without trading show that most volatility stems from trading behavior itself (French and Roll, 1986), or at least does so when we can reliably observe fundamental values in experimental markets (Smith, Suchanek, and Williams, 1988) or in the closed-end funds that trade in field markets. Volatility that stems from trading behavior rather than changes in fundamental values is excess volatility, and in the days when information processing was far more expensive than it is today, excess volatility and its concomitant trading volume was what allowed stock exchanges to exist so that financial scientists could observe and analyze the activities within. Low-cost information processing means that low-volatility, low-volume exchanges could prosper -- but it doesn’t mean that they will.

Researchers studying price volatility have traditionally focused on the volatility of a single security’s time series of prices, or a single index, rather than focusing on shifts of volatility across securities. This unfortunate but understandable focus (understandable given formerly high costs of information processing) obscures important cross-security trading dynamics. These dynamics are physically visible in futures markets, where crowds of traders will follow the action across the floor from pits trading stable contracts to pits trading volatile contracts (Baker, 1984). Cross-security trading dynamics also appear in observations of hot IPO market sequences (Friend and Herman, 1965; Ritter, 1991), where price bubbles rotate from one market sector to another -- casinos at one point, Internet stocks at another. The traders involved involved in these sequential sectoral bubbles certainly overlap in the futures markets, and likely overlap in stock markets, and the investment bankers issuing hot sector IPOs certainly overlap. An adequate agent-based market simulation will let traders choose among several securities, and a successful simulation will show traders crowding into one volatile security after another. Researchers determining the empirical statistical regularities in a single time series of prices need to recognize that these trading crowds mean that the forces driving that time series differ at different times.

Predicting these events, or at least identifying them as soon as possible, seems obviously important to large, successful traders, and seems as well to be sine qua non in a scientific finance agenda. The facts in the field go beyond just identification and prediction, though. Plentiful evidence, mostly discounted or ignored by orthodox Neoclassicals, show that large successful traders may in fact induce these events through a process called market manipulation (Abolafia, 1988; Wolfson, 1980). The market manipulator uses trading capital and public relations to create a price trend that attracts a trend-following crowd (for trend-following evidence, see Andreassen and Kraus, 1990), either by bidding and talking prices up or by selling and knocking prices down. Once the trend-following crowd begins extending the price movement, the manipulator can reverse course, using the naive traders crowding in to sell his inventory dear or cover his short positions cheap, depending on his initial posture in the security. With a conspirator, the manipulator can create the appearance of price movements without even committing capital, making so-called wash sales that change prices without any shares actually changing hands. In the US, New York Stock Exchange rules have prohibited wash sales since the 19th century, and the Securities Exchange act has prohibited market manipulation since the 1930s. Prosecutions here seem much rarer than violations, though. Regulators will accept almost any pretext for commitments to a trading strategy, plausible or not, so while some price bubbles may emerge from natural confluences of events, others seem likely to stem from intentional strategies. Other countries, such as Japan, have no prohibitions on market manipulation. Caveat emptor is the broad operating principle.

Agent-based models, though, put econophysicists in a position to explore the operating principles far more detail that just caveat emptor. The legal and practitioner literatures make it clear that thinly-traded markets in small, obscure stocks are far easier to manipulate than high-volume, highly-liquid markets in large, well-known stocks, but don’t even begin to quantify just what percentage of market turnover manipulators need to commit as trading capital to overcome any natural tendencies for prices to revert to levels rather than begin following trends. By creating predatory agents with look-ahead strategies for creating artificial trends by buying high to sell even higher to naive trend-following agents, agent-based models can explore how these predatory agents’ relative numbers and endowments affect market stability.

Of course, the regular emergence of bubbles and crashes in experimental and agent-based markets with no manipulators shows that naive trend-following agents alone are quite enough to generate a bubble on their own with no artifice. Indeed, bubbles are so frequent that one could easily question whether the trading communities in any stocks have a tendency to revert to a given level. A quick check of the financial scientist’s most reliable guide to field market behavior, the closed-end funds, makes it clear that most stocks spend most of the time quietly tracking reasonably-near the net asset value approximation of their fundamentals. Still, market manipulation can affect even these stocks with their frequent availability of the most transparent information on their fundamental values, as the practitioner literature on the great closed-end country fund bubble of 1989-1990 very strongly suggests (Wall Street Journal, 1989). These events also show that even the most determined efforts by arbitrageurs using short-sales and secondary-offerings to increase security supplies may not contain bubbles once they take off (Wall Street Journal, 1989a, Wall Street Journal, 1990). Eventually, the trading crowd shifts its attention to the next attraction and the higher proportion of capital from the value-based strategies in the ecology tends to make prices revert back towards fundamentals

Whether a trend stems from intentional market manipulation or just a chance confluence of traders, it offers a real market phenomenom that lets superior traders, human or artificially intelligent, profit from their rough predictions. In humans, this ability to predict the course of trends and trade appropriately seems to be an innate talent, rather than a learnable technique. Conversations with experimentalists who use laboratory market trading and field market trading simulations in academic and professional training suggest that about one in forty or one in fifty of the people they train display this innate talent. When these talented trainees have accepted job referrals and begun trading professionally, they have uniformly achieved successes that parallel their demonstrated abilities in lab markets and field simulations. This success over a broad range of market conditions suggest that institutions might be able to screen large numbers of individuals to find traders with the talent to earn excess returns in bubbles and crashes. To the extent that these traders respond to objective price patterns rather than hormonal phenomena, it also suggests that econophysicists could develop large, diverse populations of artificially-intelligent agents, and screen them to find the subset of agents capable of trading successfully in bubbles.

In financial terms, a self-funding econophysics should evolve specialized artificial agents fine-tuned for bubble and crash strategies, and identify humans with talents that prosper in those phases of security trading as well, perhaps even developing or finding agents with the capacity to identify securities ripe for market manipulation. Creating traders to trade in all a security’s trading phases certainly presents an interesting intellectual challenge, but offers no financial incentive in the typical phase when successful strategies can earn no more than normal returns to capital. Furthermore, at any given time, many securities trade in the most attractive trading phases of bubbles and crashes, where large potential returns for successful traders dominate market impacts and commissions. Econophysicists exploring the 40,000 or so securities trading around the world can put together large portfolios of current bubbles, current crashes, current mean-reversions, and current correctly-priced stocks. Reasonably good agents specializing in bubbles, crashes, and mean-reversions make good financial sense, because these securities offer substantial excess returns for agents to capture, if only in part. Even the best agent specializing in correctly-priced stocks fluctuating along with their fundamentals can’t expect to capture excess returns from stocks with no excess returns to capture.

As advancing Internet-based trading technologies spread low-cost desktop access to continuous auction and dealer markets to more and more trend-following individuals (Domowitz and Steil, 1999), they vastly increase the potential for bubbles and crashes in these dominant market designs. Last century, continuous auction and dealer markets rapid displaced call auctions in stock markets around the world since they increased trading volume and concommitant profits for market makers. That they increased price volatility along with volume was no concern for market makers flush with higher income. Today, new electronic exchanges and crossing networks are crowding in to offer instant bid-ask availability to rapidly increasing numbers of get-rich-quick trend-followers. Supported by continuous trading news and rumors from stock touts operating in on-line chat rooms, by day-trading firms with trading schools and easy intraday credits, these new on-line traders flock to the most active stocks, as rising proportions of small-lot trades in the ongoing Internet stock bubble show. Mixing in on-line access for algorithmically-based artificially-intelligent traders raises the potential for blindingly-fast volatility, since these silicon traders can react in a infinitesimal fraction of the response time of the carbon-based bipeds that have dominated the markets up to now. Combining technological advances with existing market structures can give us more bubbles and crashes faster.

In orthodox theory, though, arbitrage is such a powerful force in preventing bubbles and crashes that markets need no designers to keep prices closely tied to values. Econophysicists should know that current markets in fact frequently treat arbitrage as just one trading strategy in the ecology. Rietz (1998) shows experimentally that arbitrageurs who strictly enforce price limits do a great deal of work for far less profit than they might earn in other trading pursuits. Unless delivery of the exact same commodity across markets makes arbitrage self-liquidating (e.g. buying gold in London and selling it in New York), and trading conventions recognize this strategy, its practical power falls far short of its theoretical power. Even where traders in real markets can resolve delivery issues (a significant hurdle for many so-called arbitrage strategies), limits on the capital it musters relative to other, frequently naive strategies can leave prices fluctuating at quite different levels from values (as Fishman (1995) illustrates in a fascinating article about a bankruptcy arbitrage). Where regular traders can’t resolve delivery issues, as in different securities with the same "risk" (e.g., Modigliani and Miller, 1958), so-called "risk arbitrage" influences prices so little that corporate principals (who can deliver one stock for another in mergers and acquisitions) can still earn rich arbitrage profits in the market for corporate control.

As just one trading strategy in the ecology, arbitrage is so weak compared to the trend-following in bubbles and crashes that even in closed-end country funds with transparent underlying fundamentals, enormous arbitrage positions from short sales and hot IPO market issues failed to eliminate the 1989-1990 bubble. Increasing the number of Spain Fund shares outstanding by 10% with short sales didn’t eliminate the 145% September premium in that stock. Increasing the number of Germany Fund shares outstanding by 150% with a secondary issue didn’t eliminate the 100% January premium in that stock, nor did the issue of millions more shares in three clone funds for Germany. Still, that experience, and hot IPO markets in general, hasn’t kept market designers who do recognize excess volatility problems from suggesting arbitrage strategies to resolve them. Schwartz’s (1988) proposal for stabilizing arbitrage accounts funded by traded corporations has little prospect by itself of dampening the increasing volatility that could stem from the bubbles and crashes stimulated by easier trend-trading.

Nor do the old proposals to dampen excess volatility via "sand-in-the-gears" transactions taxes (Tobin, 1978) make much sense, at least in terms of a fixed tax level. For securities in the value-following trading phase, a transactions tax that discourages trading in bubbles and crashes would make liquidity prohibitively expensive, while a tax that permits value-following trading would have no effect whatsoever on prospects for traders in bubbles and crashes. Given that markets serve a real need for liquidity (a world without trade would be greatly impoverished), governments can’t afford to set transactions taxes at levels that discourage speculation. A variable transactions tax might seem like an adequate response to this objection, but that founders on the difference between properly discouraging traders who merely follow trends, and improperly discouraging traders who have real information bearing on appropriate price levels for a security or currency. Most large price moves are uninformative regarding the fundamentals outside of market dynamics, but not all of them are.

Experience with experimental call markets in the laboratory suggests that implementing Schwartz’s (2000) proposal for a reversion to call markets from continuous markets wouldn’t eliminate bubbles and crashes, although it is certainly a step in the right direction of maintaining liquidity while moving away from the excess volume, excess volatility market designs that evolved through time and prevail today. Indeed, call markets may help reduce the frequency of bubbles and crashes, if not contain them once trading behavior launches them. By concentrating the available liquidity at one point in time, rather than spreading it out over a period, call markets make it harder for small orders to move prices, and consequently harder for a series of small orders on one side of the market or another to begin establishing a trend. This last effect is, of course, mere conjecture, and needs testing in validated market experiments and simulations before applying it to our many different market designs in bonds, stocks, commodities, currencies, futures, options, and so forth.

The prospects for dampening or eliminating excess volatility are not entirely bleak, though. One need only look at the evolution of market designs that promote excess volatility to gain insights into its reduction. Foremost among these are the commodities futures markets, which have turned the stimulation of trading into high art. The exchanges that operate these markets pay particular attention to margin, adjusting it frequently to keep it at the lowest level consistent with the risks of uncovered losses for the clearinghouse. As underlying volatility in a contract increases, exchanges raise margin requirements to ensure that brokers can liquidate an account before the trader runs out of cash and begins playing with the broker’s money, or worse, the clearinghouse’s money. As underlying volatility eases, exchanges lower margin requirements to ensure that the likely daily or intraday price changes have a significant effect on the trader’s account balances. Without the excitement of spectacular potential gains, speculative volume in the contract can fall below the levels that the exchange needs to pay its costs for operating the contract. This can happen independently of the need or desire that hedgers may have to lay off their risks.

Beyond their careful attention to margin, commodities futures exchanges reveal another key aspect for controlling volatility: setting different rules for speculators on the one hand, and traders motivated by real liquidity needs, such as producers, processors, and consumers, on the other hand. Exchanges will waive margin requirements for qualified hedgers to the extent those hedgers use futures contracts to reduce price risks they already take in their business, or, for those doing business in the delivery area, to transact through the exchanges. Since exchanges are motivated to maximize trading volume and commission income, and earn no rewards for improving price discovery by eliminating bubbles and crashes, that is as far as they go in developing these differences. New exchanges that instead endeavored to minimize excess volatility, and paid no attention to trading volume (as is now possible, given computer-driven reductions in transaction-processing costs), might go much further in differentiating their treatment of liquidity trades and speculative trades.

As a first step in reducing speculative volume and its cointegrated volatility, existing market designs suggest that redesigned financial markets ought to require margin deposits equal to the full face value of the contract or security for speculative trades (which would include all trades by speculators, and non-liquidity-based trades by qualified hedgers). Decentralized markets, such as foreign exchange markets, would need central bank regulations on lending in one currency to make deposits in another. To the extent that these measures reduced existing encouragement for speculative trades, and hence speculative trading volume, they would tend to increase the information content of prices by increasing the proportion of total trading volume from producers, processors, and consumers who tend to be more informed about real supply and demand. Conceivably, exchanges might require super-margins that exceed the face value of the traded instruments, although this would stimulate off-exchange trading even more than requirments for face-value margins. The effects of increasing margins to reduce speculative trading may seem like pure conjecture, but the history of commodity markets suggests that producers, processors, and consumers felt that price volatility was far less onerous before the introduction of futures exchanges than afterward, and that politically they came within a hairsbreadth of banning exchanges as nothing more than gambling dens. It’s worth noting, also, that differentiating between speculative trades and liquidity trades in secondary markets for stocks and bonds is far more problematic than drawing that distinction in primary and derivatives markets for commodities and foreign exchange. Still, the opportunity to participate in markets with improved price discovery might persuade institutions and individuals to part with information certifying that their trades lead to a net increase or decrease in their investments, rather than simply a rebalancing within their portfolio.

It’s a technically-small and politically-huge step for primary market regulators and primary market exchanges to adequately differentiate between speculative trades and liquidity trades and then develop rules that promote accurate price discovery by applying margin or super-margin requirements to reduce speculative trading volume. The Chicago School’s "free men, free markets" movement has invested enormous significance in its adamant denial of trading behavior effects such as bubbles and crashes precisely to avoid regulations that directly address the nature of exchanges as excessively-volatile gambling dens. Still, the exchanges have already made their pact with the devil, using industry-captured market regulators to regain credibility with the public after some of the more searing experiences of market manipulation such as 1929 stock market bubble and the 1981 silver bubble (Abolafia and Kilduff, 1988). This gives the lie to the Chicago School concern that the gains from eliminating excess volatility might fall short of the losses from inviting potentially-manipulative politicians into the process. The politicians are already there, and they could do significant good by directing regulators to research and set up institutions that operate more in the public interest of discovering accurate prices and less in the private interest of maximizing trading volume.

Indeed, once politicians direct regulators to take that first step in primary markets, it puts them in a position to go beyond to steps that are technically more difficult but laden with the potential of eliminating large price bubbles and crashes, if not excess volatility altogether (White, 1996). In primary markets, they could combine transactions taxes or regulations with arbitrage funds to further decrease or eliminate the proportion of speculative trading involved in price determination. For example, regulators could use market access regulations (or a high Tobin tax) to create a closed call-market exchange where only permissible (or tax-exempt) liquidity trades drive price discovery. Once such a closed exchange discovers the price that balances fundamental supply and demand, regulators could use that fundamental price to clear an open call-market exchange that allows (or exempts) speculative trades paying full face-values for positions, with a Scwhartz arbitrage fund financing any speculative supply/demand imbalances occurring at the fundamental price.

Laboratory experiments tell us about the willingness of traders to accept a given price (Andreassen and Kraus, 1990), and careful attention from the futures exchanges themselves teaches us about the dampening effects of high margin requirements on speculative fervor. Together, these observations provide a certain plausibility for this conjectural call-market design freed from excess volatility via a directed combination of high margin requirements, Tobin taxes and Schwartz arbitrage funds. Still, regulators should engage researchers to explore many different market designs and thoroughly test them before committing to pilot projects in individual contracts and securities, much less major market redesigns for entire exchanges. This testing should include pure Tobin taxes, pure Schwartz arbitrage funds, and other existing proposals that may seem inadequate for one reason or another, because, as with the flight of the bumblebee, empirical proof trumps any supposed logical shortcomings. Thorough testing will require far more carefully-calibrated simulation and experimental platforms than currently exist, but the econophysics community has the right agent-based simulation tools, and together with the sociological, psychogical, legal/regulatory and experimental economics communities, they could calibrate those simulation tools with appropriate combinations of field and experimental observations. A shift in regulatory research budgets from their current Neoclassical orientation to a more open, empirical agenda could easily provide the funding to support a major increase in simulation and experimentation from current levels.

John Holland (1992) emphasizes that interdisciplinary, complexity-science-based simulations must establish their worth in "reality checks" by displaying evidence of the pervasive phenomena observable in the real systems that they model. For econophysicists scaling the ivory tower in finance, the Neoclassical orthodoxy says that rational, efficient markets are the pervasive phenomena in real-world finance, the Neoclassical apostasy says that broad systematic departures from expected utility are the pervasive phenomena, and both sides have developed elaborate equations and statistical methodologies for expressing and defending their positions. With a contradictory standoff in the profession over just which phenomena truly pervade, econophysicists may want to follow Agassiz’s dictum to "study nature, not books" and just skip that ivory tower debate entirely. There are many relevant if underreported observations that can help econophysicists develop a self-funding research agenda that makes real advances in predicting price movements within the existing ecologies of strategies, and in predicting how the market redesigns could change those ecologies of strategies. This paper outlines two relatively original and potentially useful approaches to trader behavior and market design, and gives a few examples from a scattered but significant literature that could provide some empirical guidelines for econophysicists seeking to establish the truly pervasive and important phenomena that they should model.

A significant advance in understanding trader behavior in existing market designs would not only enrich humanity’s understanding of its world, but would likely enrich the researchers as well. This paper contributes by emphasizing the researcher’s need to model the dynamic population movements across markets that stem from shifts in volatility -- movements that observers can still see physically in the few remaining physical trading floors as traders in commodities exchanges run across trading floors from pit to pit (Baker, 1984), or as floor brokers in stock exchanges run from specialist station to specialist station, following the rising volatility that represents volume and trading opportunities. Nonstationarity in security returns structures is far faster than the shifts in underlying corporate structures that have worried some of the researchers trying to model a single return-generating function for a long time series. Significant changes can occur inside a single trading day.

Learning how these traders shift from security to security by accurately simulating their cross-market dynamics in agent-based models holds out the prospect of finding significant predictability in security prices through the evolution of artificially-intelligent agents that consistently lead the movements by buying into bubbles and short-selling into crashes. Knowing that some human traders have an innate talent for performing the same feat, and that some of these traders prosper year after year in the zero-sum commodities futures trading game, suggests that evolving such agents is not necessarily beyond the capacity of the econophysics tool kit. Finding that significant predictability could validate the model for the finance apostates, if not for those orthodox skeptics who like to ask why you’re not rich if you’re so smart, but once you are rich, want to add a degree of freedom and attribute your wealth to luck rather than smarts. In any case, finding predictability would support ongoing research, since the addition of more artificially-intelligent agents should generate more volatility in more securities, shifting more wealth from inferior traders to superior traders, and likely raising the bar for superior trading ability.

An improved understanding of how market designs affect trader behavior offers more than just the satisfaction of knowledge and riches. Understanding market design principles could let econophysicists design new markets that substantially improve price discovery by eliminating bubbles and crashes, while maintaining liquidity for those trades that really need it. Hot IPO markets show that traders create bubbles and crashes in specific sectors fairly frequently, and comparisons of trading and news in the top percentage gainers and losers show bubbles and crashes in specific securities quite frequently, so even if market-wide bubbles and crashes are still fairly rare, acquiring the knowledge to eliminate bubbles and crashes should be an important goal.

This goal might be urgent as well. With ongoing evolution in complex adaptive financial systems, and particularly the recent emergence of technologies that let econophysicists create artificially-intelligent agents that can directly and economically access exchange-based computer crossing systems via Internet-based trading screens, and that let millions of individuals of modest means have the same access, the myriad frictions that have stabilized centuries of stock market trading history may dissolve so quickly that regulators have neither the time nor the knowledge to slow an out-of-control bubble explosion that takes markets completely out of touch with fundamentals. Happening at the same time that combinatorial technology growth introduces unprecedentedly rapid change in the fundamentals, bigger and more frequent bubbles and crashes in individual stocks and sectors could lead markets into increasingly absurd allocations of hitherto precious investment capital. Perhaps in an era of increasing surprises in technological evolution and wealth creation, markets might actually serve humanity better as devices for randomizing rather than guiding investment, production, and consumption decisions, but that is certainly a departure from their usual justification.

These conjectures on how markets will evolve with advancing technologies may seem overblown, bringing to mind a consultant trying to scare clients, then charge them for a palliative. Still, markets already have a history of hot IPO markets shows quite a lot of misallocated capital and subsequent remorse. To the extent that informed supply and demand can establish reasonable fundamental prices in a maelstrom of change, society seems to need new market designs to protect those deliberations from the undue influence of speculative crowds bearing trend-chasing algorithms. Econophysicists with agent-based models seem better positioned to fill that demand than practically any field outside of experimental economics, and some experimental economists are not only quite open to the econophysicists simulation tools, but interested in the challenges of market design (ESL, 2000). Even if regulators don’t see a need to fund that research, econophysicists who successfully predict bubbles and crashes could certainly afford to.

Abolafia, Mitchel Y., 1984, "Structured Anarchy: Formal Organization in the Commodity Futures Markets," in Peter Adler and Patricia Adler, (eds.), The Social Dynamics of Financial Markets, pp. 129-150, Greenwich, CT; JAI Press.

Abolafia, Mitchel Y. and Martin Kilduff, 1988, "Enacting Market Crisis: The Social Construction of a Speculative Bubble," Administrative Science Quarterly, Vol. 33, pp. 177-193.

Allais, Maurice, 1955, "Le comportement de l’homme rationnel devant le risque: Critique des postulats et axiomes de l’Ecole Americaine," Econometrica 21, 503-46.

Andreassen, Paul B., 1988, "Explaining the Price-Volume Relationship: The Difference between Price Changes and Changing Prices," Organization Behavior and Human Decision Processes, Vol. 41, pp. 371-389.

Andreassen, Paul B. and Stephen J. Kraus, 1990, "Judgemental Extrapolation and the Salience of Change," Journal of Forecasting, Vol. 9, pp. 347-372.

Arthur, W. Brian, John H. Holland, Blake LeBaron, Richard Palmer, and Paul Taylor (1997), "Asset pricing under endogenous expectations in an artificial stock market," in W. B. Arthur, D. Lane, and S.N. Durlauf, eds., The Economy as an evolving complex System II (Addison-Wesley, Redwood City, CA).

Baker, Wayne E., 1984, "Floor Trading and Crowd Dynamics," pp. 107-128 in Peter Adler and Patricia Adler (eds.), The Social Dynamics of Financial Markets, pp. 41-55, Greenwich, CT: JAI Press.

Chen, Nai-fu, Raymond Kan and Merton H. Miller, 1993, "Are the Discounts on Closed-End Funds a Sentiment Index?," Journal of Finance 48, pp. 795-800.

DeLong, Bradford J. and Andre Shleifer, 1991, "The Stock Market Bubble of 1929: Evidence from Closed-end Mutual Funds," Journal of Economic History 51, No. 3, (September), 675-700.

Domowitz, Ian and Ben Steil, 1999, Automation, Trading Costs, and the Structure of the Trading Services Industry," Brookings-Wharton Papers on Financial Services 1999.

Ehsan, Ahmed, Roger Koppl, Barkley Rosser, and Mark White, 1997, "Complex Bubble Persisistence in Closed-End Country Funds," Jounal of Economic Behavior and Organization, Vol.32, No.1 (January), pp 19-38.

ESL, 2000, www.econlab.arizona.edu/

Farmer, J. Doyne, 1999, "Physicists Attempt to Scale the Ivory Towers of Finance," Computing in Science and Engineering, November-December, pp. 26-39.

Fishman, Ted, 1995, "Short & Ugly: Scattered Story," November (Z9511F03.html)

French, Kenneth R. and Richard Roll, 1986, "Stock Return Variances: The arrival of information and the reaction of traders," Journal of Financial Economics 17, No. 1, September 1986, pp 5-26.

Friend, Irwin and Edward S. Herman, 1964, "The S.E.C. Through A Glass Darkly," Journal of Business, Vol. 37, October, pp. 387-405.

Holland, John H., 1992, "Complex Adaptive Systems," Deadalus, Vol. 121, No. 1, Winter, pp. 17-30.

Kindleberger, Charles Poor, 1978, Manias, Panics and Crashes: A History of Financial Crises, Basic Books: New York.

Lee, C., A. Shleifer and R. Thaler, "Investor Sentiment and the Closed-End Fund Puzzle," Journal of Finance (March 1991), pp. 75-109.

Lejeune, Robert, 1984, "False Security: Deviance and The Stock Market," pp. 173-194 in Peter Adler and Patricia Adler (eds.), The Social Dynamics of Financial Markets, pp. 41-55, Greenwich, CT; JAI Press.

Machina, Mark, 1987, "Choice Under Uncertainty: Problems Solved and Unsolved," Journal of Economic Perspectives 1 (Summer 1987)

Mandelbrot, Benoit G., 1963, "The Variation of Certain Speculative Prices", Journal of Business October, 1963, pp. 394-419.

Malkiel, Burton G., 1977, "The Valuation of Closed-End Investment Company Shares," Journal of Finance, Vol. 32, No. 2, June, pp. 847-859.

Mirowski, Philip, More Heat than Light, New York, Cambridge University Press, 1989.

Modigliani, Franco and Merton Miller, 1958, "The Cost of Capital, Corporation Finance, and the Theory of Investment" American Economic Review, June, vol.48, 261-297.

Niederhoffer, Victor and M.F.M. Osborne, 1966, "Market Making and Reversal on the Stock Exchange," Journal of the American Statistical Association, v. 61 n. 316 December pp. 897-916

Osborne, M.F.M., 1965, "The Dynamics of Stock Trading," Econometrica, v. 33, No. 1 January, pp. 88-113

Rietz, Thomas A., 1998, "Enforcing Arbitrage Restrictions in Experimental Asset Markets," University of Iowa working paper, September.

Ritter, Jay, 1991 "The Long-Run Performance of Initial Public Offerings," Journal of Finance, Vol. 46, No. 1 (March 1991), pp. 3-27.

Schelling, Thomas C., Dynamic Models of Segregation," Journal of Mathematical Sociology, 1971, 1, 143-86.

Schwartz, Robert A., 1988, "A proposal to stabilize stock prices," Journal of Portfolio Management, Fall, pp. 4-11.

Schwartz, Robert A., 2000, Building a Better Stock Market: The Call Auction Alternative, editor, Kluwer, forthcoming

Smith, Vernon L., Gerry L. Suchanek and Arlington W. Williams, 1988, "Bubbles, crashes and endogenous expectations in experimental spot asset markets," Econometrica, Vol. 56, No. 5, September 1988, pp. 1119 - 1151.

Thompson, Rex, 1978, "The Information Content of Discounts and Premiums on Closed-End Fund Shares," Journal of Financial Economics, Vol.6, No. 2-3, June-September, pp. 151-186.

Tobin, James, 1978, "A Proposal for International Monetary Reform," Eastern Economic Journal, Vol. 4 (JulyOctober), pp. 15359.

Tversky, A., & Kahneman, Daniel. Rational choice and the framing of decisions. Journal of Business, 1986, v. 59, pp. S251-0S278.

Von Neumann, John, and Oskar Morgenstern, Theory of Games and Economic Behavior, Princeton: Princeton University Press, 1944.

Wall Street Journal, "Spain Fund Shares Soar More Than 32% on Heavy Buying by Japanese Investors," by M.R. Sesit and M. Siconolfi, September 21, 1989

Wall Street Journal, "Spain Fund Falls, But Short-Sellers Can't Find Shares" by M. Siconolfi, September 28, 1989

Wall Street Journal, "As Country Funds Face Rival Clones, Will Prices Stop Climbing?" by M. Siconolfi and M.R. Sesit, January 11, 1990

White, Mark, 1996, "A New Foreign Exchange Market Design" working paper.

Wolfson, Nicholas, 1980, "Investment Banking," in Twentieth Century Fund, 1980, Abuse on Wall Street: Conflicts of Interest in the Securities Markets, Quorum Books: Westport, Ct.

Zweig, Martin E., 1973, "An Investors Expectations Stock Price Predictive Model Using Closed-End Fund Premiums," Journal of Finance, Vol. 28, No. 1, March, pp. 67-78.