The following article was written by Clive Davidson, a London-based financial journalist.

The foreign exchange market has changed dramatically over the past 25 years. Since the system of fixed exchange rates broke down in the early 1970s, stability has been replaced by a frenetic market that operates around the clock and across the globe, turning over around $1000 billion each day. But while this extraordinary growth has produced opportunities for enormous profit, factors such as the increased use of derivatives and the declining influence of the central banks means that the market has perhaps never been more risky.

In this increasingly complex and fast-moving environment, traders need tools to help them make decisions. Predicting the market has always been something of a black art. But the accumulation over recent years of huge databases of price and other information is enabling scientific analysis to be applied to market behavior. From this is emerging a new breed of computer-based forecasting and trading tools.

In a few days in September 1992, amidst the turmoil surrounding sterling's exit from the European Exchange Rate Mechanism (ERM), the Quantum hedge fund run by New York investment guru George Soros made a profit of $1 billion. In 1994, Japanese oil company Kashima Oil owned up to losses on dollar derivative deals of $1.5 billion. In between those two extremes are many cases of large gains and substantial losses in the foreign exchange market. Some involve banks - Citibank, for example, made ($1 billion: to be confirmed) profit in foreign exchange trading in 1994. Others involve corporates, like food and drinks group Allied Lyons, which lost 150 million pounds in currency speculation in 1991.

These examples highlight some of the major developments in the market over the past few decades - the globalization of trading, the huge amounts of money now flowing through the market, the volatility of prices, the pace at which events can unfold, the increasing complexity of the financial instruments such as derivatives, the power of hedge funds and derivative deals to sway the market while the influence of the central banks diminishes, and so on.

The global trading day begins in Tokyo and Sydney and, in a virtually unbroken 24 hour cycle, moves around the world through Singapore and Hong Kong to Europe and finally across the U.S. before being picked up again in Japan and Australia. Although London remains the largest center for foreign exchange trading, followed by the U.S., significant volumes are now traded in the Asian centers, Germany, France, Scandinavia, Canada and elsewhere.

London handles 30% of the foreign exchange market, with a daily turnover of $300 billion, according to the most recent major investigation of the market by the central banks and the Bank for International Settlements (BIS) conducted in April 1992. In that month, the U.S. had a turnover of $192 billion, followed by Japan with $128 billion. The Asian markets have been growing steadily. In the three years to 1992, Singapore had increased its turnover by a third to $74 billion to overtake Switzerland as the fourth largest trading center. Hong Kong traded more than Germany and France, while turnover in Spain, Greece and Denmark grew by more than 100% since 1989.

"One sign of just how international the total foreign exchange market has become is the large share of cross-border business," says the 1992 BIS report. It estimated that over half of all transactions are with counterparties in another country. But the BIS warned that its report presented only a snapshot of a fluid and changeable market. "The amount, type and locus of trading in the global exchange can vary considerably from month to month," it said.

Although the volume of currency traded can vary considerably depending on events, it is widely believed that the foreign exchange market now trades on average $1000 billion a day. This is almost a five-fold increase over the past 10 years - the average daily turnover in 1986 was $205 billion. It is also roughly the same as the total reserves of the over 140 countries belonging to the International Monetary Fund (IMF). The "fairly calm month" of April 1992 saw trading average $880 billion a day, according to the BIS. The normal unit of trade for freely traded currencies was $10 million. The average size of spot deals involving U.S. dollars was $4 million. Swap and options deals were four times as large, while the average size of Deutschmark/pound sterling deals reached $32 million.

Such huge deals are often made under conditions of extreme volatility. During some periods quoted prices can change 20 times a minute for major currencies, with prices on the on the U.S. dollar/Deutschmark rate, for example, changing up to 18,000 times in a single day. The dollar/Deutschmark rate can move 3% in 10 minutes and up to 16% in one week.

Exchange rates can be particularly volatile around major news events or during periods of crisis, such as during the ERM crises of 1992 and 1993. The British pound lost 15 per cent of its value in a few days in September 1992, providing the opportunity for Soros' Quantum Fund to make its killing. Over the longer term, movement can also be dramatic - the dollar lost 50% of its value between 1985 and 1987, for example.

Many economists expected that the lower inflation achieved in much of the developed world in recent years would bring lower volatility in the foreign exchange market, but this has not happened. A number of reasons have been suggested for the increase in volatility, including advances in telecommunications which can disseminate information instantly, the use of computerized trading systems, the presence of speculators in the market and so on.

Whatever the reason for its increase, volatility brings with it risk. But it is also a condition for trading, with only options likely to be profitable in periods of low volatility. Swings in volatility can be sharp and unexpected, as was graphically illustrated in March and April 1995, at the time when the dollar was tumbling and the ERM was coming apart at the seams.

Dollar/Deutschmark traders would probably expect the spot price to trade in an average daily range of around 80 pips (single price increments), according to Richard Irving, assistant editor of Risk magazine. Reporting on the turmoil in foreign exchange market in March and April, (Risk, May 1995, page 25), Irving says that on several occasions the daily pips range rose to five times the average.

During the same period, the implied volatility of one-month dollar/Deutschmark options soared to 19%, with the crisis having a dramatic affect on spreads as well. Spreads on one-month dollar/Deutschmark option volatility increased from 0.25% to 0.6%, with traders reporting spreads of as much as 2% for large deals. Meanwhile, spreads on Deutschmark/lira options increased from around 0.5% in late 1994 to as high as 10% at their worst in March.

There was little advance warning of these sharp moves. Trading was light in the days leading up to the onset of the volatility. The suddenness of its arrival and its extremes showed just what today's markets are capable of.

"The large intra-day swings we've seen in the dollar/Deutschmark and dollar/yen (in March and April 1995) have effectively changed people's perception of risk," says Leslie Grant, executive director of corporate foreign exchange at Swiss Bank Corporation's capital markets and treasury division. "All of a sudden, end-users have been alerted to the vicious speed with which the market can move."

The market turmoil of March and April took place at a time of growing unease at the U.S. government's failure to tackle its massive current account deficit. But this and other fundamental economic and political conditions prevailing at the time does not fully account for the market's behavior. It does illustrate, however, the declining role of fundamental factors in the foreign exchange market. In fact, in some periods the markets appear to move contrary to fundamentals. For example, when the French franc came under attack in the market in late 1992 to July 1993, indicators suggested that France's economy was in fact robust and improving. In the 1980s, as the U.S. trade deficit grew, the dollar actually strengthened.

Changes in exchange rates over the years can at times hugely exaggerate changes in the underlying economies of countries. For example, the value of sterling halved against the Deutschmark between 1981 and 1982 - from DM5 in 1981 to DM2.33 in 1982. It also halved in value against the dollar in the early 1980s - falling from $2.50 in 1980 to under $1.25 by the mid-1980s. Meanwhile, the value of the yen rose over three times in relation to the dollar over a period of 20 years - from $1:Y360 to $1:Y100. All these moves were out of all proportion to the relative changes in performance of the respective countries' economies during the same period.

One suggestion as to why the market can seem so out of synch with fundamentals is the fact that the amounts traded in foreign exchange are now around 20 times the level of actual world trade. The sheer volume of money flowing through the market creates its own dynamics. These dynamics are further amplified and distorted by factors such as the massive growth in derivatives and the use of electronic trading systems.

There has been a steady growth in the use of derivatives since financial futures were first introduced in the 1970s. The proportion of spot business to forwards, futures and options has declined in the London market from 73% spot business in 1986 to only 49% in 1992. Of the non-spot business in 1992, 6% was outright forwards, 40% swaps and 5% futures and options. Over the past few years futures and options as a proportion of the market have been growing around 35 per cent a year. Trading in derivatives such as options can be very focused and highly leveraged - the premium paid for an option is usually only a small proportion of the amount covered by the deal.

The exercising of options and other uses of derivatives during times of crisis, such as the ERM crises of 1992 and 1993, can lead to major distortions in the market. This has been worrying central bankers for some time. The Hannoun Report, prepared by central bankers from the Group of 10 for the Bank for International Settlements in November 1994, concluded: "The development of derivatives - options in particular - may effect the defense of an exchange rate target by creating hedging-related positions which may initially slow down the outbreak of crises but which may intensify them once they begin to materialize."

The distortive effects of derivatives is just one of the factors which have led to the a decline in the ability of central banks to influence the market through intervention. On more than one occasion in recent years this ability has been challenged, and at times, defeated. In the single day of 30 July 1993, the Bank of France used FF300 billion of its reserves in an attempt to keep the franc within its ERM limits. In the three months of June to September 1992, the Bank of England estimates that it used around $40 billion a month to defend sterling. Neither central bank succeeded. During the same period of 1992, Sweden spent $26 billion in six days trying to defend the krona, even raising interest rates to 500% overnight at one point, but to no avail.

The desperate attempt of central banks to resist the tide of the markets was seen again 3 March 1995, when the dollar dropped below Y95. The world's central banks mounted a concerted round of intervention not seen since the G5 countries' Plaza Accord agreement to intervene in the market to halt the dollar's inexorable rise during 1985. Unlike 1985, however, the recent operation, led by the US Federal Reserve, and involving the Bank of Japan, the Bundesbank and 15 other European central banks, failed to deliver. By April, the dollar had fallen a further 10%. This was despite a 0.5% cut in official German money rates and central bank intervention, which is estimated to have involved the purchase of more than $20 billion against both the yen and the Deutschmark. But as Oswald Grübel, general manager of Credit Suisse, observes, "The trading community can mobilize trillions of dollars to move the markets in the short term." Irving of Risk says that derivatives traders effectively neutralized the central banks' March 1995 intervention.

The trading of options only became possible when the computers necessary to perform their complex pricing calculations became widely available. Advances in technology have affected the market in other ways as well. The emergence of electronic information sources, such as Reuters' screens which first appeared in the 1970s, meant that news could be instantly disseminated around the globe. The competitive advantage of early access to information disappeared.

Screen trading was introduced in the early 1980s and has subsequently spread to the point where over 40% of foreign exchange trading in Japan goes through automated dealing systems. The proportion tends to decline as trading volumes increase. Only 24% of trading in the U.K. went through automated systems in 1992 and 32% of U.S. trading, according to the BIS report. The proportion increases in countries with a turnover less that $20 billion a day - in South Africa, for example, 90% of deals are through automated systems.

The effects of the proliferation of technology are hard to quantify. Automated trading systems were held partly responsible for the stock market crash of 1987, although lessons have since been learnt about how to build in more flexible and effective control mechanisms. But automated trading and real-time information feeds are believed to have at least contributed to the increase in volatility in the foreign exchange market. As the market has been growing in size and volatility, trading has become increasingly pressurized. Banks and other traditional trading organizations face tougher competition than ever before, with declining margins hitting profits. Insurance companies, pension funds and corporate treasuries have entered the market directly, not only to hedge their commercial investments but to attempt to trade for profit. Sometimes this has had disastrous results, as with Allied Lyons, Procter & Gamble and a number of others.

Meanwhile, so-called hedge funds, like Soros' Quantum Fund, or other big names like Tudor Investment or Tiger Management, are able to focus massive sums of money on particular points in the market. Analysts such as Michael Lipper of U.S. research firm Lipper Analytical, believe that hedge funds control up to $1000 billion in assets which they can bring to bear on the market.

In today's huge and volatile market, in which traditional factors such as economic fundamentals and central bank intervention are losing their influence, where can traders find guidance for their decision making? Almost since the markets began traders have been looking for ways to predict the movement of prices. For many years, "chartism," or technical analysis has been widely used, especially by short-term traders. Most of these techniques plot the historical movement of prices on charts and then attempt to find patterns or identify events which indicate changes in market trends or in the direction of price movements. According to an article in the Bank of England quarterly bulletin in November 1989, 90% of dealing institutions used some form of charting or technical analysis in foreign exchange trading. Two thirds said that for short-term forecasting - intra-day to one week - charts were at least as important as fundamentals. Given that intra-day traders account for 90% of the volume of foreign exchange trading, technical analysis clearly plays an important role in decision making in the market.

But technical analysis is controversial and dismissed by many as a form of mysticism. However, recent studies, such as a paper by W. Brock, J. Lakonishok and B. LeBaron published in the Journal of Finance in 1992, showed that technical analysis can at times predict market trends with greater than 50% accuracy. The authors suggest that it is "quite possible that technical rules pick up some of the hidden patterns" or that they pick up some source of risk that is not formally measured but which the market gives a value. The authors go on to say that "why such rules might work is an intriguing issue left for further studies." This lack of explanation, along with difficulties in interpretation of the charts and their apparent over-simplification of what is clearly a highly complex and dynamic process, has led to some disillusionment with the approach. In recent years, attention has turned to new technologies for pattern recognition and optimization of trading strategies. One of the most enthusiastically adopted has been computer-based "neural networks."

As their name suggests, neural networks mimic the way the human brain is thought to work. Neural networks process sets of information through a network of interconnected artificial neurons, rather than processing items sequentially as in a conventional computer. Because they have proved effective in finding patterns, for example in machine vision systems, it was thought they would quickly find patterns in the financial markets. Initial experiments were often very promising and many systems have been tried in foreign exchange trading only to yield disappointing results. For example, Credit Suisse ran a neural network from 1988 to 1993 which initially made a 9% return per annum. Then for two years it made a 9% loss per annum and was switched off. Citibank had a similar experience with a high profile neural network in its currency dealing room in London. The system was installed in January 1993 and given capital of $10 million to trade. From tests it was expected to make a return of 18 per cent per year. Within two years the system had been quietly turned off.

Neural networks, like technical analysis, are "black box" techniques. Because of the way in which they process information, it is difficult to tell how they have reached a solution. The failure of technical analysis, neural networks and similar attempts to find simplistic solutions to forecasting the market has led attempts to take a more scientific approach to the problem.

Any science requires the materials or environment in which to perform experiments to test hypotheses. The material evidence of market behavior is the data of price movements and other variables, such as volumes and the identity of traders. Although end-of-day prices have been available almost from the time the market began, it is only relatively recently that detailed data of intra-day events in the market has become available. It must be remembered that 90% of foreign exchange volume is handled by intra-day traders, so it is analysis of the daily data which is most likely to uncover the structure and forces which shape the market.

Because currency trading takes place in a vast network of dealing rooms rather than central exchanges, foreign exchange data is more difficult to obtain than, say, stock market data. By far the largest database of foreign exchange prices is that collected by Zurich-based Olsen & Associates, a major developer of software tools for financial forecasting, trading and risk management. The company's database now contains around 200 million quoted prices that have been collected since 1986 on more than 80 currencies.

Research using datasets like the Olsen & Associates database has already revealed structures and patterns in the market. For example, the intra-day data clearly shows the markets have a "seasonality" in which volatility tends to increase when more players are present - such as from 2-4 p.m. Greenwich Mean Time when both the European and New York markets are operating - and decreases when there are fewer players present, such as at 4 a.m. GMT which is when Asian traders break for lunch.

The data also reveals that the markets are fractal - that patterns of volatility measured at 10 minute intervals are similar to those measured at one-hour intervals and similar again to patterns measured at daily intervals. This fractal structure can be explained by the fact that the markets are composed of "heterogeneous agents."

"Most economic analysis of markets contains only three kinds of agents, market traders, informed and noise (liquidity) traders," says Charles Goodhart, the Norman Sosnow Professor of Banking & Finance at the London School of Economics, who believes that analysis of the newly available datasets marks a watershed in the understanding of financial markets. "It is a thesis of many that this division is woefully inadequate; that much of the dynamics of markets is made up of the interplay between these differing kinds of agents, who differ in the amount of time they devote to following separate kinds of information, and, in consequence, in their operational horizons; an hour is a long time to a market trader, a week is a long time to a politician, a quarter is a long time to a portfolio manager."

Taking these scientific findings into account, and using advanced technologies such as "genetic algorithm" techniques for searching through data, new decision support tools are being built which are designed to forecast the market more reliably. It is not only the banks and other professional trading institutions which can benefit from these new tools. As the Bank for International Settlements observed in its annual report of 1993: "Technology, innovation, free capital mobility and investors' desire for international portfolio diversification have by now all combined to increase vastly the potential for shifting large amounts of capital around the world, and across currencies, at great speed. For, even leaving aside outright speculation and also the relatively new highly leveraged 'hedge funds,' far more investors now have an interest in exchange rate developments than formerly. In other words, even the managers of traditionally conservative institutions such as pension and insurance funds, as well as retail investors, must now necessarily take account of perceived exchange rate prospects."

[June 1995]