UBS embraces machines in quest to boost corporate bond trading
(Bloomberg) -- UBS Group AG is ready to jump on the electronic-trading bandwagon to help boost its U.S. corporate-debt trading business.
The Swiss bank is testing a computer program that will price and trade some inquiries received by its bond desk -- which are typically for smaller, so-called odd lots for high-grade bonds. The bank recently hired Gokul Chebiyam, who previously founded a machine-learning corporate bond startup, to its investment-grade credit trading team.
UBS is the latest bank to embrace efforts to modernize the more-than $8 trillion market for corporate debt, where most deals are still carried out over the phone or via messaging services. Last week, Morgan Stanley President Colm Kelleher told a conference in London he can’t wait for fixed-income trading to go electronic. The market has lagged stocks where the adoption of electronic trading has been near universal.
“There is significant room for expanding the electronic-trading effort when it comes to credit across the sell side," Gary Rapp, head of investment-grade trading at UBS, said in an interview. “We can now teach a computer to process thousands of data points, which at times can make its pricing even sharper than a human. This frees up traders on the desk to have bigger picture, thematic conversations.”
Rapp joined the bank last year from Goldman Sachs Group Inc., as UBS steadily added to its credit-trading team in an effort to grab greater market share in the U.S.
Instead of getting traders to spend time trying to find the right price for all in-bound inquiries, Rapp says the algorithm is better positioned to arrive at the best price to trade the bonds. This will be mainly aimed at odd lots to start with -- a trade request where the order amount is a departure from the normal size for that security.
The machine can use several variables as inputs, such as looking at the company in question’s full selection of bonds, benchmarks and liquidity scores, and calculate the best price at which to trade. Pricing these bonds is a time-consuming task that computers can do better, according to Rapp. The bank’s goal is to eventually build a system that is capable of pricing and managing its own trading book and inventory of corporate bonds, he said.
UBS would be joining other banks who’ve been tinkering with this technology and seeking to expand it to a broader set of corporate bonds. Notably, Goldman Sachs has experimented with its Goldman Sachs Algorithm, a program that allows investors to trade U.S. corporate bonds without having to communicate with a human at the bank.
Investors are finding that electronic-trading technology has matured to the point where three quarters of credit investors say they can easily buy or sell orders up to $5 million, according to research by Greenwich Associates. That compares with about one third in 2016.
Its uptake by banks has raised questions as to whether traders will be automated out of a job. For Rapp, the challenge of trading bigger bond orders is also a reason algorithms won’t necessarily pave the way for extinction of humans on the trading desk.
“On large block trades, you still need trust and that’s where a trader with strong relationships can add a lot of value," he said.
--With assistance from Sonali Basak