Morgan Stanley Overhauls Equity Systems to Cut MicrosecondsNina Mehta
Morgan Stanley has overhauled its trading infrastructure for U.S. equities to keep pace with a marketplace in which millionths of a second count.
The bank modernized its technology for stock trading to save fractions of a millisecond and is now moving clients onto the new platform, Bill Neuberger, global co-head of Morgan Stanley Electronic Trading at the New York-based company, said by phone. A millisecond is a thousandth of a second and a microsecond is a millionth. When carrying out buy or sell orders for customers, the changes will enable the bank to get 99 percent of the shares available on exchanges at a specific price, up from the high 80s two years ago, he said.
Faster computers and regulatory changes aimed at spurring competition among exchanges have increased the speed at which stocks change hands over the last 15 years. High-frequency trading, which relies on fast access to changing data, now accounts for more than half of U.S. stock volume. Neuberger and Andrew Silverman, who run Morgan Stanley Electronic Trading, oversaw the firm’s 18-month project.
“As markets continued to evolve, the importance of speed and smart coordination became more and more prominent,” Neuberger said. “Trading fundamentally changed. We needed to think about it at the level of microseconds as opposed to milliseconds in the previous world and that involved rethinking how our trading systems operate.”
The technology overhaul is unprecedented for Morgan Stanley, said Neuberger, who has worked at the bank for 24 years including eight in information technology where he wrote software code. The firm typically made changes to its equities technology at a more gradual pace, he said. He declined to say how much the bank spent on the project.
Morgan Stanley’s equities trading unit accounts for more than 11 percent of average daily volume in the U.S., up from less than 10 percent two years ago, Neuberger said. Neuberger, who said he expects the upgrade to increase market share, and Silverman are responsible for the trading technology systems across Morgan Stanley’s entire equities division including the electronic unit.
The overhaul and focus on low-latency trading, or buying and selling with a minimum of delay, comes as banks face decreased profitability from stock trading following a decline in U.S. volume for three years in a row.
Morgan Stanley’s equities trading revenue was $1.59 billion in the first quarter, a 19 percent drop from a year earlier, excluding accounting charges. The business, headed by Ted Pick, produced the second-most revenue among global investment banks last year, trailing only Goldman Sachs Group Inc.
The firm has been the largest broker by volume traded on the Nasdaq Stock Market in the exchange’s listed companies every month since at least June, according to data from Nasdaq OMX Group Inc.
Morgan Stanley plans to buy the remaining 35 percent of a brokerage joint venture from Citigroup Inc. this year, giving it full ownership of a wealth management division with $1.79 trillion in client assets. Chief Executive Officer James Gorman said earlier this year that his firm will get a revenue benefit since it no longer has to give Citigroup a portion of retail trading orders.
The bank introduced the new trading infrastructure internally in recent weeks and is shifting clients onto it, according to Silverman, who was head of U.S. algorithmic trading at Goldman Sachs before joining Morgan Stanley in 2006. The firm will shift its U.S. options client business to the new technology by mid-year, futures and European trading by the end of the year and Asia in 2014, he said.
The overhaul encompasses trading-related systems and technology including how market data from 13 stock exchanges is processed, algorithms that execute orders, Morgan Stanley’s dark pools and so-called smart routers, or software that decides which venue should receive the smaller pieces of a larger order and for how many shares. It also includes hardware, the way Morgan Stanley connects to exchanges and how it uses its main data center for equities trading in Weehawken, New Jersey.
Morgan Stanley has three dark pools: MS Pool, Trajectory Cross and MS Retail Pool. The retail venue was introduced about two years ago.
“Reducing latency, increasing automation and maintaining proper risk measurements are major objectives of financial firms,” Anthony Dostellio, Chicago-based managing partner at Objective Paradigm Inc., a technology recruiting firm specializing in financial services, said by phone. “More organizations are facing a greater emphasis on getting data, analyzing data and reacting to data.”
Spending by banks and brokers for U.S. equity trading IT infrastructure is projected to be $2.09 billion this year, down from $2.27 billion in 2012, according to Bob Iati, partner and global head of consulting at New York-based research firm Tabb Group LLC. The costs include market data, feed handlers, high-speed messaging, data centers and networking.
As markets moved to more automated processes from exchanges centered on human traders over the last decade, brokers developed algorithms or electronic execution strategies to buy and sell bigger orders over a period of time without exacerbating price moves. Programs employed by brokers and their clients usually take liquidity by trading against bids and offers that have been placed on exchanges and dozens of alternative venues.
“You want to hit a bid but it’s made up of 13 different bids,” Silverman said, referring to orders spread across stock exchanges. “The ability to maximize the capture of that liquidity was declining.”
High-frequency traders may see incoming orders take liquidity in a stock at several exchanges and cancel their bids or offers on other venues. While firms including market makers usually quote on many platforms because they don’t know which venues investors or brokers will use, they don’t want to get more executions than they expect. Their place in the queue of orders awaiting execution at a certain price also differs across venues, giving them an incentive to quote in multiple places.
For investors and brokers, that poses a problem, Silverman said. They see quotes available across exchanges and are frustrated when they try to trade with them and the bids or offers vanish or only fewer shares are available.
“In today’s market, liquidity is fleeting,” Silverman said.
The problem of market makers canceling quotes at better prices isn’t new and occurred when trading was handled by individuals manually entering bids and offers on their computer screens. Some brokers in recent years have timed when their algorithms send trade requests seeking to access buy and sell orders at different exchanges to try to reach as many quotes as possible at the same time.
Morgan Stanley’s average hit rate of about 99 percent means that if 10,000 shares are available at a specified price, the bank could take all but 100.
The firm changed its smart router as part of the overhaul. Decisions about how to handle orders now incorporate real-time learning algorithms that evolve as they get new information from trading orders, improving the ability to get more shares by selecting venues where trades are more likely, Neuberger said.
The data center is where the smart router’s “brain” sits, with smaller order-handling systems co-located at exchanges or other venues, Neuberger said. Dark fiber is used to send orders rapidly between the Weehawken building and other sites, he said.
“Infrastructure at a bank isn’t a commodity,” David B. Weiss, a senior analyst at Boston-based research firm Aite Group LLC, said by phone. “It’s part of their business. They’ve all got to keep up with the Joneses.”
Banks employ systems engineers and software engineers and individuals with advanced degrees in computer science, electrical engineering, math and physics to build and manage trading systems, Dostellio said. The smartest and most versatile are also pursued by high-frequency trading firms, gaming companies, corporations that make consumer products sold online and social-media firms, he said.
Morgan Stanley’s electronic trading group worked in the past with application developers to construct execution algorithms and smart routers. Trading executives decided how they wanted the algorithms to behave and developers built them.
For the project the bank brought so-called enterprise infrastructure specialists into the discussion with developers for the first time, Neuberger said. These are engineers focused on hardware, network switches and operating systems. The involvement of both groups ensured that the hardware and software were integrated to “trim microseconds,” he said.
The changes will also benefit Morgan Stanley’s business of servicing other brokers, Silverman said. The bank, like its rivals, handles some trading services for small and mid-size brokers. The new technology will let it process more orders, he said. Morgan Stanley has more than 275 broker-dealer clients globally and expects that number to increase with the new infrastructure, he said.
The bank maintained two separate equities trading technology infrastructures during the overhaul. Doing so doubled the computer-coding work to prepare for regulatory changes such as the so-called limit-up/limit-down system that curbs trading in individual stocks when prices move rapidly and ordinary technical updates to systems.
“We’re betting a lot in a resource-constrained environment that this will be a big selling point for us globally,” Neuberger said. “The decay due to the changing market environment and latency issues is so fast that what you had four years ago -- it gets old.”