Battle of the black boxes

On March 3, Banc of America Securities began testing a computer program that could change the way Wall Street does business -- and cost hundreds of high-paid traders their jobs.

Called Premier Block Trading and devised by computer scientists and mathematicians at BofA, the program automates a process at the heart of institutional trading: principal deals, in which brokerages put up their own capital, thereby assuming risk, to facilitate large-scale, or block, equity trades for mutual funds, hedge funds and other big investors.

As it is now, BofA's clients have to phone one of the bank's traders to get a quote on a principal block deal; soon they will be able simply to pull up a Web page and enter a ticker symbol for the stock they want to buy or sell and the number of shares. In an instant PBT's computer algorithms will factor in the particular stock's trading characteristics and BofA's own position in it and then generate buy and sell quotes. Clients will be able to execute trades of up to

$20 million in value apiece with the click of a mouse.

"It's a very valuable tool," says Kevin Connellan, director of trading at $490 billion-in-assets Northern Trust, who has been beta-testing the system. One day late last month, Connellan traded blocks of 450,000 and 79,000 shares of two New York Stock Exchange­listed stocks within seconds on PBT. "My main alternative would have been down on the NYSE floor, but there I'm only getting 4,000 or 5,000 shares before the price moves. This just gives you a clean way to get your trade done."

BofA's PBT is scheduled to go live later this month. As revolutionary as the bank's block-trading system may be, it is only the latest in a wave of computer-centered innovations that are dramatically transforming the multitrillion-dollar institutional trading business. Electronic communications networks, such as Archipelago Exchange and Instinet, have been steadily siphoning volume away from conventional stock markets. Private trading networks, like Liquidnet and Investment Technology Group's Posit, now allow institutional investors to trade large blocks -- tens or hundreds of thousands, or even millions, of shares -- directly with one another. Big brokerage firms have been eagerly embracing technology designed to automate most trades. Tellingly, Goldman, Sachs & Co. let go 35 institutional traders and salespeople in February. Even the hidebound NYSE -- under new CEO and former Goldman president John Thain, a Massachusetts Institute of Technology graduate -- is moving toward automating the human share-auction system that has held sway there for more than 200 years.

Not even technophiles expected automation to entrench itself this far, this fast. The phenomenon portends enormous changes for Wall Street. Up for grabs is not only the balance of power between the technocrats and the traditionalists within brokerage houses but also the balance of power -- and business -- among firms. Trading is a core function. Firms that have the brains and the bucks to build the best black boxes will gain a significant edge in institutional trading, which in turn will benefit virtually all of their other activities, from investment banking to proprietary trading (committing capital for the firm's own account rather than to execute a customer's order).

The majority of institutional trades today are already executed in so-called low-touch fashion, either by a fully automated system or by a human trader employing similar technology. In 1995 program trading, in which baskets of stocks are bundled together by computer for cheap, efficient execution, accounted for a mere 12 percent of Big Board activity. Today nearly half of NYSE volume consists of program trades.

Big Wall Street firms -- notably, Goldman, Morgan Stanley and Credit Suisse First Boston -- have been pioneers in developing (or acquiring) and implementing automated trading. Groups within the firms operating separately from traditional trading desks use mathematical algorithms embedded in computer software to mimic human traders. Institutional clients access these algorithms directly from their desktops and control their trade executions -- essentially renting the brokerage's technology. The simplest algorithms, known as smart routers, break blocks into smaller pieces and route them to whichever exchange or alternative trading system offers the best price. More sophisticated algorithms allow for different levels of aggressiveness and risk tolerance and support complex, cost-minimizing strategies.

Faceless software programs such as Pathfinder, RediPlus and Passport now preside over a trading world where larger-than-life figures like Cy Lewis of Bear, Stearns & Co., Gus Levy of Goldman and Jay Perry of Salomon Brothers once dominated with a combination of skill, nerve and luck (a booming voice didn't hurt either).

"Firms have been investing a ton of money in intelligent routing and other low-touch technology," says William Cline, head of the capital markets practice at consulting firm Accenture. "It reflects the reality that in today's markets, the power of the computer is much better suited to handle trade-offs between price, immediacy, anonymity, size and so on than humans are."

The BofA block-trading system dares to venture into a domain that is regarded as a last preserve of human traders. Even the most ardent advocates of computer trading tend to concede that the largest and most complex orders require a trader's blend of discretion, intuition and responsiveness. After all, a firm's capital is at risk. Can bundles of silicon and plastic be trusted to execute a block trade in which a firm acts as a principal, rather than as an agent seeking out counterparties in the open market? BofA's system purports to do precisely that. If successful, it's bound to be imitated.

To be sure, human traders aren't likely to completely vanish. Technology still can't react to Federal Reserve Board announcements and other breaking news (though it's getting close). And some clients may never completely trust their most sensitive trades to a black box, no matter how advanced its technology. Still, the traders who survive will have to adapt. "If you don't know how to use electronic trading as a tool," says Ciaran O'Kelly, head of stock trading at BofA, "you're not going to make it."

The relentless drive toward all-out automation has a telltale side effect: fewer traders. The job cuts at Goldman, which trades 450 million shares per day around the world through its RediPlus platform, are far from unique. Over the past three years, CSFB has laid off 70 percent of its Nasdaq market makers and 30 percent of its block traders handling NYSE stocks. The firm's Advanced Execution Services unit processes fully 40 percent of its equity order flow -- and it is staffed not by traders but by computer scientists, mathematicians and a handful of salespeople who get clients connected to the technology. "This is the future of the business," says Manny Santayana, head of sales and marketing in the Americas for the AES group. "There will be even fewer people on trading floors."

Fewer people means fatter profit margins, of course. And that's a great relief for brokerage firms. Historically a lucrative activity, the institutional brokerage business has become a barely profitable quagmire after a decade of regulatory upheaval and competitive pressure. But with their minimal operating costs, low-touch systems can charge lower commissions than conventional trading desks -- 1 or 2 cents per share versus 4 or 5 cents -- and still deliver respectable earnings for firms. "The notion that institutional trading is a low-margin business is a myth," says one equities executive. "I make more on 1 cent business electronically than 5 cent business and a human trader, because my costs are nil."

More is at stake than just beefier margins. Firms that can't keep pace with technology will forfeit institutional order flow -- and that could jeopardize their standing in the industry. Money managers are under intense pressure to reduce trading costs, so they are directing more and more of their trades to the handful of firms that offer the cheapest execution. Those with the best automated systems will win.

"We're looking to continue to consolidate our brokers," says Richard Block, the aptly named head of trading for $240 billion-in-assets Putnam Investments. "The quality of their electronic trading platforms is becoming a much more important criterion for us when we evaluate who we trade with. It makes sense, as these systems get smarter and smarter, to use them for a bigger portion of our order flow."

Brian Fagen, a Morgan Stanley managing director who works with institutional clients to maximize trading efficiency, concurs: "You're limiting yourself as to what kinds of business you can do without having these tools. You have to offer products across all execution segments if you want to capture any kind of significant market share."

More important, a thriving institutional brokerage business is vital to a securities firm's overall prosperity. U.S. mutual funds, pension plans and other institutional investors together hold about $6 trillion in stocks and account for the vast bulk of stock market volume. Their order flow in and of itself may not be particularly remunerative for firms, but it helps to drive earnings in higher-margin businesses like investment banking and proprietary trading.

"This is about more than just trading," says Accenture's Cline. "The overall competitive position of these firms is very much at stake."

How are other businesses affected? For one, firms win lucrative stock-underwriting mandates partially on the strength of their trading desks, which can support new issues and give corporate clients important feedback about how their shares trade. No less important, order flow gives proprietary traders a wealth of data about the stock market's direction so they can take shrewder gambles. "We would expect proprietary trading in equities within the 'winning' brokerage firms to increase," wrote Sanford C. Bernstein & Co. analyst Brad Hintz in a research report earlier this year on firms' efforts to automate trading.

The rise of low-touch trading has a significant serendipitous benefit as well. It girds brokerage firms for the increasingly real possibility that regulators -- or perhaps even customers -- will force them to "unbundle" the services that those customers pay for with trading commissions. Of the 5 cents per share charged by traditional trading desks, 2 cents cover the actual transaction costs, and 3 cents pay for analyst reports, newsletter subscriptions and sundry other items. The commissions used for nontrading items are referred to as "soft dollars" on Wall Street.

Low-touch systems dispense with soft dollars. They charge customers only for trade execution. That could prove to be a timely strategic advantage for brokerage firms. "With this system we're ready for hard dollars if they come," says a trading executive for one bulge-bracket firm.

WHY ARE COMPUTERIZED SYSTEMS elbowing out institutional traders? If it's any consolation, the putative obsolescence of the traders can be blamed largely on the radically changed nature of liquidity in the stock market. In the 1960s and '70s, institutions poured money into stocks, supplanting retail investors as the main source of trading volume. Block-trading desks became fixtures of brokerage houses. Market-savvy traders worked their contacts to find counterparties for large orders. They also committed their firms' capital to fill customers' orders instantly, then skillfully unwound those positions over time to avoid losses, or even to reap gains.

But the old order has been turned upside down. Market-structure reforms intended to help investors ended up fragmenting trading volume among several market centers and drastically reducing the number of shares available at the best prevailing prices. In 1996 the Securities and Exchange Commission, responding to a price-fixing scandal involving Nasdaq market makers, passed order-handling rules that effectively encouraged the migration of quotes from Wall Street brokerages to electronic exchanges. Soon the ECNs were executing roughly half of Nasdaq's volume. Bid-ask spreads shrank, shriveling further in 2001 when Wall Street, under pressure from Congress, began quoting prices in decimals rather than in fractions of a dollar. As the spreads that traders profited from all but vanished, they committed less capital. And decimalization allowed NYSE floor traders, small-time trading shops and even individuals to trump large orders by simply offering to buy or sell as few as 100 shares for a penny better than the best quoted price. That made it far more difficult for institutions to move big blocks.

Meanwhile, during the past decade or so, new classes of investors such as index funds, statistical-arbitrage hedge funds and other active traders began exerting greater influence over the market. Index funds are huge users of program trading. Stat-arb and other hedge funds seek to profit from smaller, frequent trades that capitalize on tiny, often fleeting anomalies in the statistical relationships between securities. Decimalization has enhanced the effectiveness of these strategies, because it provides more price points. But the strategies have also contributed to rapidly flickering price quotes and the evaporation of liquidity at the best quoted prices. The average size of a trade has declined dramatically since 1996, from 1,400 to 492 shares on Nasdaq and from 1,489 to 450 shares on the NYSE.

These developments unsettled institutions. In the past two to three years, they have been aggressively seeking alternatives to conventional brokers and exchanges for trading large blocks (Institutional Investor, April 2002). And Wall Street has responded by turning to intelligent technology to retain buy-side flows.

"Decimalization has driven a rise in intraday and statistical-arbitrage strategies," says Rohit D'Souza, head of global trading at Morgan Stanley. "The trading styles of these participants, combined with the widespread use of program trading by index and active funds, has dramatically changed the composition of liquidity in the market. Therefore, traders need different types of tools to access that liquidity. And by adopting these tools, they can significantly reduce their execution costs."

The most basic of the new electronic tools, and the first to be implemented by brokerages, is "smart order-routing" technology. Smart routers are powered by relatively simple algorithms that scan the various exchanges, ECNs and market makers where shares of a given stock might be available, find the best prices and route orders accordingly. Smart routers were first developed for day traders, and adapted for institutional use a few years ago (Institutional Investor, June 2000).

Firms such as Hull Group; Spear, Leeds & Kellogg; and Timber Hill Group were among the first to make a business of providing smart routing technology to institutions. Goldman bought Hull in 1999 and SLK, which built the RediPlus system, the following year. The current version of RediPlus is a blend of the two firms' technologies, plus additional features that have been added in the past few years. The platform now sits on 2,000 client desktops, and its volume grew by 55 percent during the first four months of 2004 alone. Goldman won't say what percentage of its overall institutional flow goes through RediPlus, and SLK is maintained as a separate broker-dealer from Goldman's traditional desk. But for perspective, consider that the 450 million shares traded on average on RediPlus every day around the world are nearly one third of the 1.5 billion traded daily on the NYSE.

Timber Hill, now called Interactive Brokers Group, is one of several such firms, including ITG, Instinet Group and UNX, that are unaffiliated with diversified securities houses and provide electronic agency brokerage services. "You can program your computer to do whatever you do," says chairman Thomas Peterffy. "That's why we don't have traders. It's all done with computers." Interactive Brokers' workforce of 450 includes sales, back-office and technology support staff in the 40 markets globally where it executes orders.

But most big Wall Street shops aren't ready to completely automate. Take Morgan Stanley, another early pioneer in electronic institutional trading. In 1997 the firm began developing a system it calls Passport, an electronic portal for institutional clients to various automated execution products. As with Goldman's RediPlus and other sell-side trading platforms, clients can connect to Passport through software provided by Morgan Stanley or through a connection with their own order-management software. Institutions can use Passport to access Morgan Stanley's smart router or to send baskets to the program trading desk. Clients also can enlist the help of a series of Morgan Stanley developed algorithms that oversee the execution of orders according to a number of cost-minimizing techniques, which the firm calls benchmark execution strategies, or BXS. Like offerings from CSFB, Goldman and some other firms, Morgan Stanley's BXS includes an algorithm that delivers the volume-weighted average price for a given stock, as well as one designed to limit so-called implementation shortfall. That's the difference between the result a trader could get upon immediately executing a large order at the prevailing price and the actual result achieved over the time it takes to execute that block. In the old days it was impossible to pursue such strategies without handing the order over to a human being.

But Morgan Stanley also gives Passport's technology to its sales traders, who are now far more productive as a result. In fact, Passport started as an effort to help the firm's traders cope with the fragmenting effects of the 1996 SEC order-handling rules. Whereas once a single trader would handle ten to 15 stocks, now he can manage 40 to 50, and Morgan Stanley makes markets in 1,500 Nasdaq stocks, up from 500.

According to UBS stock analyst Glenn Schorr, Goldman and Morgan Stanley commanded an impressive 46 percent of the equity trading revenue that went to the seven largest publicly traded U.S. institutional brokerage firms during this year's first quarter. So it's little wonder that the rest of Wall Street has been desperate to catch up with these firms' early-mover advantage when it comes to trading technology.

The most successful thus far has been CSFB. Three years ago the firm shifted Daniel Mathisson, who joined CSFB in April 2000 after heading trading for quantitative hedge fund D.E. Shaw & Co., to oversee CSFB's fledgling low-touch efforts. Now Mathisson's AES group employs 30 people globally, and its computers execute the trades on 40 percent of the shares flowing through the firm. CSFB's dwindling ranks of human traders also use the technology.

JUST AFTER LUNCHTIME ON A WINTRY THURSDAY afternoon, Mathisson's is the only voice within earshot on CSFB's hockey rink­size trading floor as he explains what's happening on the AES platform.

"See, 500 just went off right there," he says, as the number of shares to buy on a 200,000-share customer order is reduced by 500 on the busily flickering, flat-panel screen. "It's entirely automated." Over the next several hours, CSFB's algorithms will execute more and more pieces of that large order until it's completely filled, hewing to a complex cost-minimizing strategy the client laid out in advance. And not one human being on the trading floor will touch the transaction.

The rest of the Street is further behind. Some firms, such as BofA and Citigroup, are moving aggressively into electronic trading. Others, like Merrill Lynch & Co., are going slower. "It's difficult to get large, existing trading teams to look at automation," says Interactive Brokers' Peterffy. "They basically fear for their own jobs."

In late 2002, BofA acquired Vector Partners, which had developed a series of sophisticated algorithms used for both single-stock and program trading. Then, in January, the bank formed a dedicated electronic trading services division, which is co-headed by veteran program trader Peter Forlenza and Ross Stevens, who has a Ph.D. in finance and statistics. Forlenza also heads the firm's equities business; Stevens serves as its COO. One month later the firm acquired Direct Access Financial Corp., a maker of software that allows institutions to access BofA's algorithms. In April the bank entered a partnership with Microsoft Corp. to research and develop more-advanced trading technology. Of the roughly 30 people in BofA's ETS group, two are traders. The rest are programmers, sales and support staff.

"The technology allows us to do more with less," says Stevens, a former quantitative analyst and portfolio manager for Goldman Sachs Asset Management. "In the old days a trader maybe could do a few million shares a day. Now, with the technological tools we have, a trader can do ten times that. To the extent that we can put these tools in clients' hands, the scale is infinite."

William Geyer, a former head of U.S. equity trading for Barclays Global Investors, is setting up a discrete electronic trading group at Citi. (The firm wanted to call the unit Alternative Execution Services, but because CSFB already has its own group dubbed AES, Citi is using the shortened Alternative Execution). The new entity will consolidate Citi's program trading, direct access, algorithmic trading and other alternative products into one system that clients can access simultaneously -- the same way that RediPlus, Passport and AES work.

"We're bringing together all the alternative execution products that surround the traditional cash trading desk," explains Geyer. "Clients want a single point of contact. This group gives them one, and helps us better prioritize product development."

Another large commercial bank, J.P. Morgan Chase & Co., recently formed a similar unit called Electronic Execution Services. And last summer UBS hired former Island ECN executive William Sterling to build out its electronic trading capabilities. In March, Bank of New York, which runs a large soft-dollar brokerage, acquired Sonic Financial Technologies, a popular provider of low-touch technology. Lehman Brothers and Merrill also are believed to be developing their own electronic platforms, possibly with cooperation from third-party vendors.

While the rest of the Street is racing to catch up, early adopters are busily trying to maintain their leads. Goldman's SLK unit, for instance, aims to update RediPlus at least four times each year, adding increasingly intelligent features. The latest release, rolled out to clients last month, incorporates a pretrade cost-analysis tool called the Guide and a pairs-trading application that hedge funds can use to put on long-short and merger-arbitrage plays. The Guide lets clients evaluate the potential effectiveness of various algorithms before choosing one with which to execute an order. The firm's 15-person Goldman Sachs Algorithmic Trading group is constantly developing and refining these algorithms, which have names like 4Cast and Piccolo, each designed to implement different tactics to minimize costs or handle certain types of stocks. On the staff of this unit are several former NYSE floor traders, who help mathematicians and computer scientists design algorithms that better reflect the realities of how trades are done.

"Two years ago it was fine just to have a basic smart router," says Duncan Niederauer, co-CEO of SLK. "Now it has to be more intelligent than that."

CSFB has introduced two innovative features to AES. One, called Storyboard, provides institutional clients with feedback about the progress of orders and market news that might affect their strategies. "You're now 60 percent of the volume in this stock" is an example of a message Storyboard might display in a client's order management system. "It's the kind of stuff that a human sales trader might call you with on the phone," says CSFB's Santayana. The other is called Portfolio Hedging Device. Mathisson touts this tool as the first to allow investors to bundle several stocks for execution as a basket while still maintaining the single-stock execution parameters that have been available in the main AES system. "We've been working on the math behind some of this for close to two years," he says.

Even relative latecomers are wielding innovations that could turn the Street's competitive balance on its head. BofA's PBT system, the existence of which was first reported by this magazine (Institutional Investor, May 2004), is one example. If successful, PBT could enable BofA to steal a march on larger competitors, particularly because as a commercial bank it has a bigger balance sheet with which to support principal trading than do pure securities firms like Goldman and Morgan Stanley. Likewise, Citi may be late in building an electronic trading portal, but it's exploiting the opportunity to build in coveted features like pretrade analytics from inception rather than having to add them later.

Automation also has fueled program trading's exponential growth. Institutions have traded baskets of stocks at the same time for many years. But when these orders first started hitting the Street in the late 1970s and early 1980s, floor brokers executed them manually, and the process could take hours.

"Basically, the floor brokers were turned into human DOT machines," says Deutsche Bank program trading head David Baker, referring to the NYSE's designated order turnaround system for routing trades to be executed. "They would take a bundle of individual order tickets, held together by a rubber band, and manually execute the trade around the floor."

During the past five years, firms have invested in technology that lets clients enter basket orders electronically, routes the various pieces where they need to go and reports back to the customer within seconds. Once limited largely to indexers and funds that were changing managers, program trades have become a lower-cost alternative to single-stock trades for active managers. "That's permanent," says Baker. "And it means the way resources are deployed is changing. Traders no longer need to allocate time to the mechanics of executing and settling a trade. Instead, they're focused on strategy."

WHO THE WINNERS OF THE BATTLE OF THE BLACK boxes will be is still anyone's guess. Some firms are further along than others, but the move to electronic trading is still young, and the technology still evolving rapidly. What's certain is that this fight's implications for Wall Street are both serious and manifold.

For one, automation is a way to reclaim profitability in an increasingly commoditized business. The cost of processing trades through automated systems is a tiny fraction of the outlay for legions of traders with salaries, bonuses, benefits, expense accounts, and so on.

But the move to a low-touch model for equity trading also coincides with the convergence of businesses within Wall Street firms. Equity divisions increasingly look like fixed-income and derivatives groups and are being headed by executives from those arenas. Goldman's equity business, for instance, has shed traders and salespeople under the influence of Lloyd Blankfein, the fixed-income, commodities and currencies guru who was promoted to president last year. James Forese, who became Citi's equity chief earlier this year, used to head the firm's bond business. Lehman's head stock trader, Patrick Whalen, cut his teeth at derivatives house O'Connor & Associates. Many of the Street's program-trading executives also grew up in the derivatives business.

Some, like Bernstein's Hintz, explain this phenomenon by theorizing that firms are moving to a more fixed-income-like business model for equities. The handful of firms that do the best job of automating will gain a greater share of institutional flows, which will help drive proprietary-trading profits. Perhaps, then, it's no coincidence that Goldman and Morgan Stanley, the most advanced firms when it comes to electronic equity trading, are universally seen as the Street's most skilled prop-trading houses. Although betting their own money on fixed-income, commodities and currencies has driven earnings at both firms of late, their most recent earnings reports suggest that revenue from proprietary stock trading is also on the rise.

That's at least one reason human traders won't disappear entirely. If proprietary trading continues to expand, people will have to oversee those operations. As for the traders who execute customers' orders, opinions about their fate vary. Some believe that computers will be able to mimic most, but not all, human behavior. "No one's been able to design an algorithm yet that can position you following a Fed announcement, stock-specific news, or the release of economic numbers," says Putnam's Block. "We shouldn't underestimate the value of information that a good sales trader can provide."

Others say that computers can, or will soon be able to, handle any trading situation. Humans won't vanish altogether, these folks say, only because some clients prefer dealing with a person who can hold their hands when things go wrong.

"The human being can't really do anything differently," insists Interactive Brokers' Peterffy. "He can just appease the person he's talking to as to why he's getting a poor execution. You know, 'I'm sorry. Next time I'll make it up to you, and by the way, let's go out and play some golf.'"