The debate over computer-assisted trading: who is right?

The debate over computer-assisted trading: who is right?

Of the many types of computer-assisted securities trading, program trading may be the most misunderstood. Here, a researcher explains what the technique is, how technology supports the trading method, and what it has done to the U.S. and international securities markets.

While there are many types of computer-assisted trading, no type has excited such interest and controversy as what has become known as "program trading." A United States Presidential Commission has looked into it. Numerous official reports, including one by the U.S. Securities and Exchange Commission (SEC), have been produced about it. And--especially during the months following the crash of October 1987--it has been the subject of media attention.

For a period of time following the crash, certain forms of computer-assisted trading were prohibited from using the New York Stock Exchange's (NYSE) order delivery systems; although this ban has now been removed, the NYSE has imposed permanent restrictions on computer-assisted trading activity.

Setting the record straight

The scale of the controversy is perhaps all the more surprising in that there are remarkable inconsistencies between what different people mean by the term "program trading." Reports in the U.S. and U.K. general media indicate that some journalists believe that program trading denotes any computer system which appears to carry out an actual trading decision on behalf of the trader. This interpretation of the term is wholly erroneous; no such systems exist.

Another widely used--but incorrect--definition of the term denotes a computer system which interfaces with a decision support system. Such a package gives the trader not only the information he might need to decide whether or not to trade, but also specific guidelines that could be interpreted as a computerized "suggestion." The most clear example of such a system is one featuring limit-marking. Limit-marking is the use of preprogrammed parameters, relating to stock prices, which cause a computer to "alert" the trader (such as by a red light flashing on the screen) when a certain parameter has been reached. This would be the signal to the trader to trade on that particular stock or other financial instrument. Limit-marking technology is already a common feature of the U.S. financial trading arena, and it is rapidly becoming popular in London and Tokyo. It is not, however, reasonable to call such a utilization of computers in financial trading a program trade.

The only accurate definition of program trading occurs when a trader aims to trade a large group, or basket, of different stocks together. The trader buying the basket of stocks will usually have a good idea of what stocks are in the basket, but will by no means know all the details of the basket's contents.

This is the true meaning of the term "program trading." It originated in the 1970s, when trades of entire portfolios came to be seen as being traded in programs. The role of computers in this type of trade is primarily to act as a system of logging the trades that have been completed. Computers are useful because a typical program trade may contain several hundred stocks, and to trade this manually would be labor-intensive and time-consuming.

Program trading imposters

There are two financial trading strategies that make use of computers and which have often been called, erroneously, program trading. These are index arbitrage and portfolio insurance.

Index arbitrage is a trading strategy by which a trader aims to make a profit by exploiting small differences, at a given time, between prices of stocks and derivative investments on the same index or exchange or--more often--on different indices or exchanges. Index arbitrage nowadays usually takes place between traditional stocks or securities and futures contracts in these stocks. The computer would identify opportunities for making profit through arbitrage, and would alert the trader to these possibilities. However, the decision to make the trade would still be made by the human trader.

Portfolio insurance involves a trader using a variety of trading strategies in order to forestall heavy loss in a portfolio he is managing, if the price of a particular stock or basket of stocks falls. One version of portfolio insurance is the stop-loss trade, whereby a certain proportion of a particular stock is sold as the price of the stock falls past certain trigger points. For example, if an investor holds 50,000 shares in Corporation X and each share is worth, say, $10, the investor clearly stands to suffer a heavy loss if the price of the stock falls to $5. This being the case, a stop-loss trading strategy might involve selling, say, 10,000 shares each time the price of the stock fell by $1. It is easy to see that such a strategy could be linked to a computer-assisted trading system. The system could even alert the trader to the fact that it was time to sell a certain quantity of a stock held in a portfolio. Computers are particularly useful where the portfolio contains several hundred stocks, all of which might have their own stop-loss requirements.

Another form of portfolio insurance involves using options or futures contracts in order to hedge against market movements in the underlying stocks for these contracts, where the underlying stocks are already held in a portfolio. Again, the speed and variety of price movements in the markets under scrutiny mean that a computer is required to undertake a rapid analysis of trades, markets, and market movements in order to ensure that changes in stockholdings are matched by changes in the holdings of options or futures contracts that act as the hedging holding.

Both index arbitrage and portfolio insurance are important types of trading strategy that can make use of computers. However, neither ought to be described as a program trade. That privilege should be reserved exclusively for basket or portfolio trades.

The right time for a program trade

Under what circumstances would a program trade be appropriate? The answer is whenever there is a large number of individual transactions to be enacted which would be cumbersome to clear individually through the market. A program trade could be a consequence of a decision to restructure a portfolio following the appointment of a new fund manager; the decision to index a fund; or the decision to effect a change in the fund's asset allocation.

One of the biggest problems associated with a major portfolio is that the restructuring usually involves dealing in bargains in a size well outside that readily accepted in the market. Not only could it prove difficult to deal in such quantities of shares, but, were the transactions not handled discreetly, the sudden activity in the shares in question would be likely to affect prices considerably. Unless a program trade is conducted, a trader wishing to execute a trade of a large portfolio of stocks would have to choose between either spreading the deals between one or more brokers to whom discretion is given and monitoring, perhaps on a sample basis, how well they were doing or, alternatively, insisting on being consulted on each and every deal, which might mean the trading organization making some 300 phone calls.

A program trade thus offers two main advantages to a fund manager: it transfers the risk--be this market, transaction, settlement, or foreign exchange-related--from the client to the broker, and it provides the client with immediate execution at a guaranteed price and with reduced transaction costs.

As might be expected, a financial trading firm interested in conducting a portfolio trade will bid for a basket only if it has a realistic chance of making a profit on the trade. Obviously, a major factor in determining the bid price is the perceived risk profile of the portfolio. Program trades are usually undertaken blind; that is, the broker does not know the exact composition of the portfolio and must use his judgement in deciding whether or not a particular portfolio should be acquired. The portfolio holder would normally disclose various characteristics of the portfolio, such as the portfolio's total value; number of securities which it contains; distribution of the securities by market and by sector or index; type (in the U.K., alpha, beta, or gamma shares); industry group and concentrations; average position size; distribution of position sizes; average trading volumes; and capitalization weightings. All this helps to assess the risk of the portfolio and, therefore, the trading firm's potential reward.

Who are the fans of program trading?

It is difficult to reach any precise conclusions regarding the extent of computer-assisted trading within the New York, London, and Tokyo financial markets. This is partly because firms involved in program trading do not readily publicize the fact, and also because such figures as do exist relate not only to true program trades but also to index arbitrage and portfolio insurance. Most trading firms we approached, when probed for information on the extent to which they engaged in program trading, declined to comment on the matter.

However, our research shows that program portfolio trading is currently almost an exclusive preserve of U.S. traders. According to the NYSE, between January and September, 1987, an average of 15,000 program trading orders passed through the Exchange per day, and on October 19, 1987, a total of 61,000 program trades passed through the NYSE on that single day. However, it must be stressed that the NYSE's definition of "program trading" includes portfolio trades, index arbitrage, and portfolio insurance.

A source at the London stockbroking firm of James Capel told us that he believed there was currently about one program trade (i.e., portfolio trade) per week carried out in the U.K. Another source at a Japanese trading firm in Tokyo said that not only Japanese trading firms but also those involved in planning corporate strategies in financial institutions were "very much interested" in both index arbitrage--and the implications which the crash of October, 1987, had for index arbitrage--and portfolio insurance.

The lack of opportunities, at present, for index arbitrage and portfolio insurance trading activities in Japan is probably due to the relative unsophistication of the computer trading frameworks in place at the Tokyo Stock Exchange. However, once index arbitrage and portfolio insurance strategies become a feature of the Japanese financial markets, they are likely to be practiced with enthusiasm.

Although international program trades have been a feature of the U.S. financial trading scenario for several years, they have only recently begun to be highly visible to money managers. With more and more pension fund managers seeking global returns, rather than only returns on domestic markets, program trades have increased in frequency and size. It is nowadays not unusual to hear of program trades involving investments of $50 to $300 million in indexed or active portfolios across as many as 16 countries. The success of these international program trades (almost invariably arranged by a U.S. trading firm from a U.S. base) has meant that an important and cost-effective tool for portfolio managers has evolved to accommodate global asset allocation.

Expert commentary, and a premonition

A financial Times conference, "Technology in the International Securities Market," held in London on March 24 and 25, 1988, prompted a broad discussion of decision support and computer-assisted trading in world markets.

Far from blaming computer-assisted trading for the events of Black Monday, Robert Gartland, managing director in London of the finance, administration, and operations wing of Morgan Stanley International, said that when an "avalanche" of 604 million shares valued at $7.6 billion had changed hands on October 19, 1987, computers prevented Wall Street from being "buried in paper," and from a consequent "unimaginable chaos."

He said that on October 19 portfolio trades, index arbitrage, and portfolio insurance had accounted for 12 percent of daily transactions on the NYSE, whereas, prior to October, 1987, they had accounted for between one-quarter and one-third of all transactions. Gartland added that although restrictions had since been imposed by the NYSE on index arbitrage and several major firms had limited such activity, it was inevitable that as automation and standardization increased the efficiency and capacity of the clearance and settlements systems, thus reducing transaction cost, the application of the emerging technologies would make it economically feasible to identify and generate new automated trading strategies at profitable margins that may not exist today.

"The net effect," he said, "will invariably be an increase in transaction volume to new record levels and a corresponding increase in liquidity, although it is likely that this trend will be concentrated in the issues of highly capitalized and creditworthy corporations and institutions." Moreover, as automated trading promotes liquidity, hedging, and arbitrage, added Gartland, these factors "may tend to moderate volatility over time rather than exacerbate it."

Another issue was raised by consultant Adrian Norman, managing director of Direct Business Satellite Systems (DBSS). He observed that fund managers and market makers were all installing similar systems. And, so, any one significant change in the markets had more or less the same effect throughout the markets. With computers all programmed in similar or near-similar ways, he said, most systems would, on occasion, work together to accentuate trends in the markets rather than smooth them out.

However, Norman pointed out that computer-assisted trading and settlement had allowed certain types of trading to be profitable, whereas without computers the trade would not have been worth the trader's while. He pointed to index arbitrage as an obvious example of this, where arbitrage gains were only possible if trading decisions could be executed rapidly.

One of Norman's most remarkable achievements was his prediction, in April of 1987, of the October, 1987, market crash. The prediction was contained in an article, "Vulnerability of Computerized Financial Trading," published April, 1987, in a number of journals.

Norman's article presented his thesis that today's global trading systems and markets were vulnerable not only to "computer failure" but also to "computer success." Norman argued that if computer-assisted trading systems were to operate as their operators intended, a crash of the market in which the systems operated would be the likely result.

Norman referred to the sudden fall of 6 percent in less than an hour in the NYSE All Share Index on January 23, 1987. At the time, Norman noted, commentators did blame program trading for this sudden fall. They alleged that computer systems run by major financial trading institutions had been reading the signals from the marketplace and that these institutions had tried to rebalance their own and their clients' portfolios by buying and selling financial instruments.

Regarding portfolio insurance strategies, Norman commented that some systems had "trigger levels" at which so-called "stop-loss orders" were issued to prevent the institution from being left holding assets of too low value. As each institution reached its own stop-loss point, its selling activity would push the price down past another institution's marginally lower level. That, Norman said, was what positive market feedback was all about.

According to classic market theory, he continued, when prices decline, buyers enter the market to pick up the bargains. He pointed out, however, that the classic model of "negative market feedback" only applies to slow-moving or almost static marketplaces. In a more dynamic marketplace, such as the world's major financial markets, traders would endeavor to sell just as the decline starts and try to buy back once they deem that the price has dropped to its lowest level. As the cost of dealing is relatively low, the cost of the selling and buying operation is small compared with the gains from selling high and buying back low. Not only would traders have every incentive to do this, Norman argued, but the deployment of computerized trading systems would mean that deals could be made in minutes or even seconds, and the price adjusted promptly.

With many systems in the market now coupled through telecommunications networks to each other and to sources of information about market movements, and with each system also knowing the rules that the other systems are going to follow, each system, he suggested, would be standing "ready to pounce," if the signals from the marketplace indicated that action was needed.

"Am I right," Norman continued, "to predict that sometime in the quite near future we shall see the crash of 1929 not in weeks but in milliseconds?"

Norman concluded his article by asking whether it might be possible to put in place a regulatory mechanism which could control the behavior of an electronic marketplace, when that marketplace could "migrate anywhere around the globe" to evade controls.

Might the central banks, or perhaps an association of the major global banks, he asked, establish some kind of "self-denying ordinance" outlawing computer-assisted trading at speeds which preclude human intervention? He suggested that this central authority might agree on some stops--triggers set by price movements beyond agreed "normal" levels--which would lead the systems to pause and allow human intervention. Alternatively, Norman asked whether "we need some kind of buyer/seller of last resort that will always intervene to prevent violent price fluctuations."

Not only did Norman forecast the crash of October, 1987, he also appears to have forecasted the way in which regulatory bodies would respond to computer-assisted trading after the crash. Much of what he predicted has now happened.

Familiarity breeds approval

Generally, our research tended to show that the more traders in the U.K. feel uneasy about computer-assisted trading, the less they have been directly involved with it. For example, we found that the only U.K. trading firm that appeared to see program trading simply as a useful trading technique rather than as a source of potential problems was James Capel, which is one of the very few U.K. financial trading firms that is actively participating in this type of trading.

In the U.S., the situation was very different. Without exception, traders we interviewed in that country said that computer-assisted trading in general, and program trading in particular, had been a great boon to them.

Ronald T. Slivka of Salomon Brothers has raised another issue, that of using a program trade to control the actual costs of making a transaction involving a portfolio of shares.

"When a portfolio is rebalanced, new money is put to work, or an equity-cash shift is contemplated," he explained, "the problem of minimizing the total cost of the transaction arises. This cost is often conservatively estimated in the U.S. at 1 percent to 2 percent of assets involved. Because of its proven effectiveness, program trading should be considered a cost control tool by money managers whenever asset shift or portfolio restructurings are contemplated."

Moreover, other services associated with program trading are often equally valuable, he said. Among these are computerized processing of trades and comprehensive reporting of trade status on a daily basis; assistance with currency transactions to meet securities settlements; pretrade analysis of equity and currency markets and of individual securities; and exploration of alternative strategies using index futures and options.

The general consensus among traders we consulted is that computer-assisted trading is beneficial, not harmful, to markets. Even the belief that program trading makes some contribution to the volatility of the markets is by no means shared by all traders. Why, then, we might ask, is it that when the book on which this report is based went to press (June 20, 1988) the eight major players in the U.S. computer-assisted trading scenario were maintaining a self-imposed temporary halt to computer-assisted index arbitrage on their own accounts? The first three, Merrill Lynch, Shearson Lehman Hutton, and Goldman Sachs, made the announcement in January, 1988. The other five joined them in May: Salomon Brothers, Morgan Stanley, Paine Webber, Kidder Peabody, and Bear Stearns.

After the decision had been announced, we asked Lou Margolis, managing director of Salomon Brothers in New York, why his firm--hitherto such a fervent supporter of index arbitrage--had made the decision it had made.

"We stand by our belief in index arbitrage," he explained, "and our faith in its future viability. We inflicted a self-imposed suspension on computer-assisted index arbitrage on our own accounts in the face of intense public pressure, but I would expect the suspension to last a matter of months, not years." He added, "All claims I have seen that this temporary withdrawal from index arbitrage has lessened market volatility are based on inadequate evidence."

This widespread suspension appears to be a major setback for computer-assisted index arbitrage, but we believe it is less significant than at first appears. It is easy, on first entering the palatial splendor of a U.S. financial trading firm's offices, to believe that traders occupy a privileged world of their own, far removed from the general public or the newsstands. But this is not the case. During our researches, it became clear that financial traders were extremely susceptible to public opinion. This should not be surprising, since the investing public is the source of the trading firms' business, and, ultimately, their profits.

In any event, index arbitrage has one supporter at the very highest level. This is Alan Greenspan, chairman of the U.S. Federal Reserve Bank. Among other testimony which he delivered on May 19, 1988, to the House of Representatives Subcommittee on Telecommunications and Finance, Greenspan said: "Insufficient arbitrage can be a deestablishing force in a declining market."

With such an influential supporter, one feels that, whatever the public pressures, the long-term future of computer-assisted trading is assured. This article is condensed from the second chapter of Computers in Financial Trading by James Essinger [C] 1988 by Elsevier Advanced Technology Publications, 52 Vanderbilt Ave., New York, New York 10017).

PHOTO : A dowry piece, silk embroidery on natural linen, Uzbekian, early 19th century From the

PHOTO : people of central Asia comes this complicated design, most likely made as a wall hanging,

PHOTO : bedspread, or curtain for a dowry. The detail reveals intricate embroidery, often done in

PHOTO : sections by relatives and then sewn together to form one piece.