Click for next page ( 83


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 82
Electronic Automation at the New York Stock Exchange CHRISTOPHER KEITH AND ALLAN GRODY In mid-October 1987 the New York Stock Exchange (NYSE) was deluged by a percentage increase in volume of orders greater than anything that had occurred in the great 1960s volume surges that created the major "back- office crises" of that period. Happily, with some creaks and groans, the NYSE's systems withstood the unexpected strain, which was far larger than the NYSE's peak load forecast. This case discusses both the evolution of those systems and some of the major issues in their development. There are two crucial aspects of almost any system effort: (1) the system as a pure technology (getting the technology to perform on its own as a system); and (2) the system as a component of a broader environment (in- terfacing it correctly with its users). These might be likened to the two problems facing the would-be inventor of an artificial heart machine. Problem one get the little valves to work properly (that is, oxygenate and pump the blood). Problem two related but very different-get the body to accept the device in the first place (whatever its supposed solution might be). At the NYSE, problem one was formidable, exacting, and full of the unexpected. Even so, it was in the area of problem two, the system in relation to its context, that the NYSE encountered its most demanding and distinctive challenges. Accordingly, in what follows, the system as a component of its environment will receive special emphasis. The Trading Process Ignoring for the moment the Exchange's automation programs, the flow of floor trading takes place more or less as follows: 82

OCR for page 82
ELECTRONIC A UTOMATION AT THE NEW YORK STOCK EXCHAI!iGE 83 First, orders arrive at the Exchange at "booths" situated on the floor periphery. The booths are maintained and operated by brokers, either "up- stairs" brokerage houses or independent "floor" brokerage firms that work on a commission basis for a number of upstairs brokerage houses. Orders may be either transmitted from upstairs systems to printers on the floor or phoned in. In either event, floor brokers take the orders received at their booths to the appropriate "post," where the stock in question is traded. The NYSE has its own central switch, Common Message Switch (CMS), which routes orders to the floor from a variety of users and systems and routes reports back to these same systems. The bulk of this traffic is routed to specialists and the posts, but some is routed to the booths. This is referred to as the OrderlReport Delivery subsystem. Second, at the post, the specialist for the stock in question and the "crowd," other brokers interested in transacting that stock, are waiting. The specialist operates in one or more of the following roles according to circumstances: (a) as a broker's broker, taking the order from the floor broker and executing it, when appropriate (for instance, in the case of limit orders); (b) as a dealer, obligated to make a two-sided continuous market, and to "bid" if no rea- sonable contra-side is available (although having to better any contra-side in order to "dealer; (c) as a trading arbiter/manager (auctioneer), responsible for a fair and orderly market; and (d) as a matchmaker/ consultant, providing a variety of informal services that help a broker to find the other side. The order, once brought to the trading post, is executed or, if it is not at current market, is left with the specialist for representation. If it is executed, the execution may be with another broker in the crowd, or with the specialist representing either his or her own quote (as dealer) or with another order left with him or her previously (as broker's broker). The functions that support this activity directly are called the Trade Support subsystem. The crucial environmental factor here is that trading must take place more or less instantaneously in the trading environment, and requires some effort from those trading. Yet the systems must also allow accurate subsequent processing of the trades made. Third, once a trade is executed, a member of the NYSE staff, a reporter positioned to "overhear" the trade, reports its elements by stroking an optical sense card and feeding it into a reader. This report goes to the Market Data System, which generates the ticker output, the High Speed Line. The reporter has a similar function in reporting quotes. Subsystems associated with these functions are called the Market Data Reporting subsystem. Fourth, after execution (and reporting by the NYSE staff reporter), the broker returns to his or her booth and ensures that the execution is reported back (sometimes by telephone, often by terminal) to the entering party. If the specialist represents one side of the execution, he or she reports similarly.

OCR for page 82
84 CHRISTOPHER KEITH AND ALLAN GRODY The NYSE facilities (particularly CMS) are generally used for specialists' reporting and sometimes for brokers' reporting purposes. This, too, is a function of the OrderlReport Delivery subsystem. Fifth, the various parties to the trade then report their portions of the trade to the clearing corporation, which compares these reports. Discrepancies, if any, are usually resolved there. Once the trades are matched, they are netted and settled. The parties to the trade deliver either the requisite securities, perhaps using the Depository Trust Company, or the requisite money. These comparison, clearing, and settlement processes represent the major components of the After-Trade Processing subsystem and in times past, the major paperwork bottleneck. In addition to enabling comparison, clearing, and settlement, this system must accommodate the interfacing needs of de- positories, transfer agents, registrars, varying state laws, and Employee Re- tirement Income Security Act regulations. A Cooperative Activity Even the rather cursory overview of trade processing above suggests that the consummation of a trade, and hence its processing, involves the coop- erative efforts of multiple independent parties. Of the four major subsystems introduced above, the NYSE, from a systems operation perspective, controls only Market Data Reporting from the input process through final service generation (generation of the ticker and the high-speed lines). In all the others, an independent entity either is responsible for input submission or, at least in part, "operates" the service offered. One prime environmental difference, then, between the NYSE's automation system and most others is the complexity of the partnerships needed to produce a result. The NYSE does not enjoy the position of a hierarchial organization with an implied line of authority to discipline all parties. Nor is there a single, distinct provider/user line controlling interfaces at all points. The NYSE's systems must serve a number of partners, each of whom has a different mix of functions and priorities. Even institutions within the same category can have very different perspectives; for example, exchange staff, specialists, $2 brokers, floor brokers, upstairs brokers, or independent brokers with widely differing policies and mixes of business. However, there is another difference that can hardly be passed over; that is, the nature and characteristics of the typical "automated." Elsewhere, automation may involve automating pro- duction workers, shipping departments, or payroll departments operations with little individual power. On the NYSE floor, however, the "automated" is likely to be an independent businessperson or person of some wealth and prominence in the community, and quite possibly a member of the NYSE's Board of Directors. Over this pluralistic universe the NYSE staff, the "automater" in this

OCR for page 82
ELECTRONIC AUTOMATION AT THE NEW YORK STOCK EXCHANGE 85 case, can perhaps impose some level of automation standards, but the ability to impose is more the ability to persuade. For a new system to work reasonably well, a substantive consensus is needed among a small but mixed community, few of whom report to a common source of authority. In a sense the Exchange floor might be compared not to a typical de- partment or business function but to a coral reef whose members share a common agreement over basic rules, but who probably have very different ideas about the future, its needs and imperatives, or the priorities and pro- cedures of the present day. Like the coral reef, any new arrangements of the Exchange's system will depend in large part on the persuasiveness and per- sistence of the members themselves. At the NYSE, members had to contribute significant time to efforts ranging from problem identification and prioriti- zation, through persuading other members of the need for change, to the choice among possible solutions, and the implementation of critical elements in the system. The Marketplace Marketplaces are notoriously difficult, it seems, for nonpractitioners to understand, especially automation experts schooled in traditional applica- tions. At first, the Exchange floor and the trade-processing function might appear to be a "plant" producing transactions. But that is only part of the trading arena's service. In addition to consummating trades for those who enter orders, the Exchange provides vital services for those who have no direct connection with it, and who do not use its services explicitly. Like any marketplace, it not only "transacts," it also "sets value." The state of the Dow Jones Industrial Average, for instance, is taken by much of the public as a measure of the financial and commercial state of mind of the country. Businesses who neither buy nor sell on the NYSE significantly guide their business plans by the values set on their own stocks, their in- dustry's average price movements, or a potential raider's relative stock value. The establishment of value in both how it is determined and the fact that it is taken as valid throughout the financial world-has profound con- sequences. The Exchange produces not only the trade and its associated information trail but also an accepted mechanism for establishing value or price. In Exchange argot the "price discovery" mechanism is itself an im- portant product. What price means is not what someone ought to be willing to pay for a stock, or might be willing to pay, but what someone is willing to pay for it now. To be meaningful, however, a price must first be ac- tionable. To be actionable, there must be buyers and sellers, each willing to transact in a given volume at a certain price. This may seem self-evident, but the result is that an exchange floor must be looked on as a kind of date

OCR for page 82
86 CHRISTOPHER KEITH AND ALLAN GRODY bureau, where highly dispersed and diverse customers with similar interests can find each other at the moment they desire. What this translates into, in technical terms, is respect for a process. It is not just what the market support system produces as a result that is conse- quential; it is, to a much greater extent than elsewhere, the process by which the result is produced. The question frequently arises, "The current processes by which the Exchange floor operates often appear cumbersome, even an- achronistic; why is it they are not replaced with something really efficient?" Or, expressed another way, "Why isn't everyone upstairs sitting at efficient computer terminals?" What such questions seem to imply is that efficiency (particularly paper processing efficiency) ought to be the priority. What else is data processing for, if not to be efficient? One should also ask, "Should the 'price-discovery' mechanism be compromised for more 'efficiency'?" Even in the interest of supposed efficiency, the environment of the Exchange moves only with great caution in tampering with the evaluation process. Perhaps the reader will see why in the next section. The Valuation Process Of the valuation methods so far devised, there is little question in the minds of most professionals that the fairest process for setting prices is the Call market. There are numerous versions of the Call market, but in all versions the various participants each of whom has a disposition to buy or sell, in differing quantities, depending on price-meet to reach a consensus price. In some markets this is called (and unfortunately so) the "fix" (see Figure 11. The NYSE uses a variant of the Call market at market opening. The trouble with the Call market is that although it serves admirably in establishing a price, it does not handle the almost equally important variable of time. The pace of the modern commercial world has become such that continuous pricing is an absolute essential for many types of instruments. Both of the two principal types of order, the market and the limit, involve at least implicit settings of both price and time. The market order specifies a fixed time (now), with price as available. The limit order fixes price (as indicated), but allows time to vary. A reasonable successor to the Call market might be one that offers both price and time advantages-a Competitive Dealer market (see Figure 2), in which the combined judgments of dealers theoretically simulate the Call market and back up those judgments with actionable prices. There is much about such a system that appeals to the intuition. For instance, if one were selling one's car, it would be reassuring to know that there would be relatively continuous competition on the issue of price. One imagines oneself shopping

OCR for page 82
ELECTRONIC AUTOMATION AT THE NEW YORK STOCK EXCHANGE 87 at various dealers over many days and then taking the best timing and price offer. BRIEF HISTORY OF TRADING The emergence of a Competitive Dealer market was the first major evo- lution that took place at the NYSE. Just after the Civil War, a broker who had not yet fully recovered from a broken leg settled down on the floor and decided to "specialize" (the stock was Western Union). He was not a spe- cialist in the current sense of the word; he was a competitive market maker- no one awarded him an exclusive franchise. By the 1930s, there were such competing "specialists" in more than 300 of the NYSE's stocks; and as late as the 1960s, there were competing specialists in more than 50 stocks. The Unitary Specialist Market Gradually, the competing market system was replaced, de facto, on the NYSE by its current Unitary Specialist market. The Competing Dealer market seems plausible, but has a flaw. In such a market buyers and sellers do not transact directly; they transact only with market makers, and they pay an interposition cost which is equal-considering first a buy and then a sell- to the dealer spread. Returning to the analogy of selling a car, it would be clear there is a potentially better alternative than finding the best dealer- and that is finding the ultimate buyer directly. Providing that the cost of finding the customer is low and that one can be sure of a fair price, the buyer and seller can then share the dealer's spread. This is the central rationale behind the evolution of the NYSE to its current structure (see Figure 31. When activity rates rose, particularly for the larger, better known companies, the chance of finding the other side directly was no longer a theoretical possibility, but a practical and achievable objective. Now at the NYSE, for example, the specialist operates as a dealer in less than 20 percent of the shares traded. Instead, the specialist helps the broker find the other side, sometimes charging a commission for this service, some- times not, but always keeping any commission below the spread. A simple computation, using easily verifiable statistics, places the value of this re- duction in interposition costs at more than $5 billion in 1986 alone a much greater savings than the cost of the entire NYSE, with specialists' profits included. Using its current market structure, the NYSE became the world's dominant equities exchange, and one whose value-setting mechanism had unequaled credibility in global financial circles. The NYSE tape print became an as- surance of an equitable deal, or at least a best effort to that end, throughout the world. This being the case-or at least the NYSE's impression of the

OCR for page 82
88 KEY: _ BUY SELL DEALER 1 1 FIGURE 1 Call market. Buyers and sellers transact directly at call price. FIGURE 2 Competitive Dealer mar- ket. Buyers and sellers transact through dealer. CHRISTOPHER KEITH AND ALLAN GRODY Continuously Arriving Orders FIGURE 3 Unitary Specialist market. Specialists act either as brokers (bringing together buyers and sellers) or, if buyers and sellers cannot be matched, as dealers. ~3 IT __:~i. Continuously Arriving Orders '. ~ C, i,[~

OCR for page 82
ELECTRONIC AUTOMATION AT THE NEW YORK STOCK EXCHANGE 89 case it is no wonder that the NYSE is reluctant to make fundamental changes in the price discovery process, no matter how tantalizing the seeming efficiencies. An understanding of this priority is essential to an understanding of the NYSE's automation. First, the institution's value determination meth- odology must be preserved, with efficiencies achieved only within that . . . Overr~c ding constraint. How automation developed within these constraints is instructive. We will introduce various subsystems in the chronological order that they were first put into practical use. The subsystems were determined essentially not by their final purpose in a grand system, but by the environments in which they operated and the human-related constraints imposed on them by those en- vironments. That is what led to the chronology. Market Data Systems The NYSE's earliest use of electronic technology dates back to 1867, when the Morse Code Ticker was created. The technology was the telegraph. The pace of trading was such that there was no particular problem in keeping up with transactions. The system's constraint was the capacity to read symbols at the receiving device and little more, other than maintainability. This simple, successful system imposed no demands on its environment. In 1881 the first electromechanical-mechanical board was installed on the Exchange, displaying last sale. The basic theme of ticker upgrades, starting as early as 1890 and continuing to the present, was speed. The 1867 ticker ran at a stately 50 to 60 characters per minute. In 1930 the ultimate seemed to have arrived, the first "high speed" teletype system with a capacity of 500 characters per minute. By 1953 developments in magnetic drum recording made it possible to add automated voice announcements. Quotations, recorded directly from the floor, were automatically relayed to subscribers who dialed the appropriate number for any of 200 active stocks (eventually expanded to 3001. This was many times faster than the manual quotation system, which dated back to 1928 and depended on Exchange operators to read the quotations from boards and relay them to member firm subscribers by telephone. In the late 1950s many new approaches including the use of television, teletypewriters, and electronic rather than telegraphic printers were studied by the Exchange. In 1961 the Exchange inaugurated the Special Bid Ask Ticker Network, making possible the transition of quotations on 800 stocks to electronic interrogation devices used by member firms throughout the United States. That same year, the Exchange's work with Teletype Corpo- ration resulted in a technological breakthrough a new design for a high- speed printer that could relay between 500 and 900 characters a minute. To this point, what had paced technology was essentially only technology.

OCR for page 82
To CHRISTOPHER KEITH AND ALLAN GRODY As Me pace of trading grew, technicians applied available technology to increase speed and improve accessibility to the market. But the joint NYSE- Teletype Corporation system in 1961 reached a new kind of barrier. Nine hundred characters per minute was about as fast as anyone could read; faster tickers would only be illegible. This led to the NYSE's first major real-time computer system the IBM- based Market Data System (MDS), a development begun in 1962. What was most remarkable was the attention paid, for its day, to fallback recovery and to the first technical attack on the 900 character limitation. When the pace of market trading approached the 900 character barrier, the system auto- matically began abbreviating, using a kind of shorthand, a technique that has been in the arsenal of ticker system designers ever since. In addition, a much more readable price display unit was developed the 900 character moving ticker display. This new device utilized luminescent plastic disks that were flipped by air jets as they moved along at speeds matching those of the 900 character ticker. Early in 1965 a new system, the Quotation Service, became operational, constituting the first major application of the new MDS. Subscribing member organizations could dial a four-digit number for any of the more than 1,600 listed stocks. Less than a second later, a spoken message was heard giving the latest bid-ask and last sale information. A message containing the open, high, low, last sale, and volume was also available. Each message was automatically composed for prerecorded words by a special voice assembly unit in the Exchange's Computer Center. Designed especially for the Ex- change, the MDS Quotation Service could handle as many as 400,000 tele- phone inquiries each day an increase of 60 percent over the previous peak load then available. The MDS was designed to handle a trading volume (then thought quite impossible) of as many as 60 million shares each day. Inauguration of the Exchange's comprehensive market price index, calculated by the MDS's computers, followed in July 1966, the same year in which optical card readers' mark-sense cards began driving the ticker. Referring to the MDS, The New Yorker in its June 1965 "The Talk of the Town" said: The New York Stock Exchange has recently automated its stock quotation services to brokerage offices, the purpose of which is to give by voice over private telephone wires the pertinent current statistics on any listed stock that the inquiring broker wants to know about.... One should not confuse the Stock Quotation Service with the more familiar ticker which records completed trades by teleprint or on paper tape rather than by voice.... The voices used to be those of a covey of girls who sat at a switchboard on the fifteenth floor of the Exchange's building at 11 Wall Street; now the voice in all cases is that of a young IBM engineer named Robert Rea, whose voice is stored in a drum inside a computer on the third floor of the building.

OCR for page 82
ELECTRONIC AUTOMATION AT THE NEW YORK STOCK EXCHANGE OrderlReport Delivery 91 In 1966 the Odd-Lot Automation (OLA) program began with the ostensible intention of increasing efficiency, improving service, and providing for an- ticipated high-volume levels in odd-lot (less than 100 share lots) trading. Odd lots were not (nor are they now) traded under the NYSE's agency auction principle, which fosters direct meeting of buyer and seller. Instead, odd-lot orders were executed by special dealers (odd-lot houses) who executed trades for odd-lot customers at the "next" execution price, plus or minus the odd- lot differential of one-eighth. The NYSE had other objectives in mind as well. By switching odd-lot orders by computer, it reduced its dependency on a manual pneumatic tube system still in place at that time. The NYSE intended to switch the odd-lot orders to an area below the floor, so it could save precious floor space. The project was initially limited in scope. What was little understood at the time was that the project had effectively ushered in the era of automated order and report delivery. For the first time, a "receptive enabler" was available with sophisticated computer capability that could be integrated with the Exchange's own systems. By 1969 the OLA was completed, providing manual underfloor pricing of odd lots for all posts and, in the process, the most extensive standardization undertaken in the securities industry until that time. At the end of the project, the securities industry and the NYSE had something that would prove more valuable than the elimination of the tube system-a set of uniform formats for teletypewriter messages in a user environment still very suspicious of any talk of computerized order delivery. Experiments and changes in the odd-lots arena did not really threaten the main, high-volume, profitable core of most key players' businesses. Not surprisingly, the successor to this project was again an odd-lot effort, now switching odd-lot traffic to the one remaining odd-lot house, Carlyle and Decoppet, off-site. The system called Odd-Lot Switch (OLS) was no sooner cutover, in 1972, than the third odd-lot project began. At approximately this time, the NYSE and the American Stock Exchange (AMEX) combined their data processing departments into a new corporation, Securities Industry Automation Corporation (SIAC)-owned two-thirds by the NYSE and one-third by the AMEX. What AMEX brought to the new corporation, among other things, was an ambitious on-line project based on a Collins front-end, and a Univac back-end. The project, Amex Computerized Order Delivery and Execution (AMCODE) was one of those far-seeing, vastly ambitious efforts that contemplated supporting a broad range of data pro- cessing services for a decade. The system, already under development, was taken over by SIAC and became a component of an SIAC plan then under development, Centralized Exchange Network Trading and Unified Reporting

OCR for page 82
92 CHRIS TOPlIER KEITH AND ALLAN GRODY (CENTAUR), even more ambitious and sweeping than AMCODE. The AM(-OnF nrniect was renamed CENTAUR Message Switch, and again was geared initially as an odd-lot system. The CENTAUR plan ultimately collapsed but the system would survive with a somewhat more modest name using the same acronym, CMS, for Common Message Switch. (Remember all that stationery with the acronym CMS?) CMS added to the standardization of order and report communications by accomplishing a remarkable technical feat. It interfaced with more than 40 firms on their terms. That is, it appeared to all these firms as a terminal on their networks. In a technical tour de force, CMS was brought on line to all 40 firms with which it interfaced in a single night, after having run in parallel with OLS for more than six months to ensure that it duplicated OLS results in all cases. What the NYSE now had was ostensibly an odd-lot system, interfacing with all of its principal customers and using standards that would apply equally to round lots. But it was an order-gathering tool of such competitive significance although still seen as an odd-lot system that the Securities and Exchange Commission, during its active pursuit of a National Market System in the mid-1970s, talked of nationalizing it. What CMS turned out to be although significantly upgraded in capacity and since migrated to Tandem computers was the major component of the NYSE's Order/Report Delivery subsystem through which all electronic equity orders to the posts (and much to the booths) now pass. Trade Support Most of these early attempts at automation had been, on their own, at least modestly successful. If these systems did not achieve everything they set out to do, they at least worked and provided some benefits. However, beginning in the early 1970s, pricked by the national attention caused by the back-office crush of the late 1960s, and spurred on by aggressive automation projects among the member firms themselves, the NYSE (and its subsidiary SIAC) made various attempts to "automate the floor," to support floor trading using automation systems. These were uniformly unsuccessful. Typical among them was Automated Trading System (ATS), which purported to provide electronic support for the specialist- in effect, becoming the first electronic book. Unfortunately, an insufficient portion of the order flow arrived on the floor in electronic form to make any such system practical. Added to this were an overly ambitious objective, too little understanding of the environment, and the lack of a strong "champion" who would defend and overcome the new experi- ment's awkwardness. The result was that ATS remained only a pilot project, and slipped from

OCR for page 82

OCR for page 82

OCR for page 82

OCR for page 82

OCR for page 82

OCR for page 82

OCR for page 82

OCR for page 82

OCR for page 82

OCR for page 82

OCR for page 82