Traders vs. Bots Moves to the Corporate Bond Market

Will machines take over the bond market too?

Banks including Credit Suisse Group AG, Goldman Sachs Group Inc., and Morgan Stanley are all making bets in that direction, unleashing new trading software systems in recent months to pick up share in the $6 trillion market for investment-grade corporate debt.

In part because of its complexity and relatively low liquidity, that corner of Wall Street has been one of the more resistant to technological changes sweeping trading, lending and just about everything else in banking. A recent study by the Bank for International Settlements estimated that only 40% of investment-grade corporate bond trading was executed through computers rather than over the phone, compared with 75% in Treasury-debt trading, 80% in stocks and 90% in a broad array of futures contracts.

Now, investment banks are pushing to stoke more electronic trading in the market, especially with small trades that otherwise might fall through the cracks. It is yet another example of banks turning to technology to try to generate revenue growth at relatively low cost.

Starting with smaller trades, those under $1 million, is a safer route for the banks because it doesn't threaten the profit margins on the big trades they do with institutions.

While the banks are taking it slow, the smaller trades eventually may be a precursor to fully automating some bigger trades, which are often trickier to execute and require more human hand holding.

"Banks see an opportunity to take advantage of trading opportunities they were missing in the past," said Richard Schiffman, head of open trading at MarketAxess Holdings Inc., operator of an electronic bond-trading network.

Some past efforts to change the patterns of bond trading have met resistance or petered out. Platforms offered several years ago by Wall Street players including Goldman and money manager BlackRock Inc. were shelved or integrated into other efforts.

But more recent initiatives, such as networks connecting bond investors to trade with each other, have started to show traction.

Credit Suisse launched its system, called CSLiveEx, earlier this year, making the decision to fully automate the smallest trades in the U.S. market for bonds of corporations with top credit. That means the trades have no direct human involvement.

In corporate bonds, traders working the phone still dominate the market, in part because there are so many various bonds to follow, all with different permutations such as maturity and yield. Some corporate bonds also don't trade that often after their initial sale, even for major companies like Apple Inc. In that type of market, human salespeople who can wheedle buyers and sellers still add value.

"We are running with the machines, rather than letting them do the running on their own," said Samik Chandarana, a veteran credit trader who was recently named head of analytics and data science at J.P. Morgan Chase & Co.'s investment bank. J.P. Morgan and its peers, which tend to have bigger balance sheets and trading desks, have invested in technology as well, primarily to aid human traders.

While automated service is more prevalent in trading for retail bond investors, and has been technically feasible for several years, it is rarely used to respond to electronic trade requests from institutions such as pensions and hedge funds.

But the increasing amounts and sophistication of data is starting to change that. And banks are eager to make money out of the thousands of small electronic trade requests they get daily.

Credit Suisse says its new technology will help the firm serve more clients without adding extra traders or bond inventory. CSLiveEx is "designed to have as many customers as possible, without using a lot of balance sheet," said Julian Pomfret-Pudelsky, a fixed-income algorithmic trader at Credit Suisse who began working on the project in early 2016.

Before launching CSLiveEx, Credit Suisse responded to fewer than 10% of clients' electronic trade requests for the smallest investment-grade corporate bond trades. Now the bank says it is responding to virtually all of those inquiries and executing trades twice as often as it did before.

Goldman in 2015 began making a limited number of trades under $1 million via software called GS Algo, and has this year expanded that to some 7,500 different individual bonds. The securities firm says the move has dramatically improved the rate at which it responded to client electronic trade requests.

"A natural place to start in the automation process is where it would enhance our ability to service our clients' liquidity needs," said Amy Hong, head of market structure strategy for global credit at Goldman. The bank has said its recently slumping fixed-income division is focusing on expanding its client roster and improving relationships.

Morgan Stanley has also recently begun fully automated trading for smaller investment-grade corporate bonds, as a way to handle some of the 10,000-plus electronic trade requests it gets a day for such trades, people familiar with the matter said.

If algorithms can handle the smaller trades, some believe banks may eventually unleash them on larger ones. Investors may also decide to break up their big orders into smaller pieces, as is common in the stock market.

"Banks are giving the algorithms a lot of at-bats," said Michael Sobel, president of TruMid Financial LLC, an electronic bond trading network. "I fully expect that the folks who are successful will keep dialing up the sizes."

Write to Telis Demos at telis.demos@wsj.com

(END) Dow Jones Newswires

October 18, 2017 07:59 ET (11:59 GMT)