Some banks have announced sweeping cuts to their fixed-income units in recent years in the face of shrinking profits.
Not Goldman Sachs. The New York bank has avoided such significant layoffs, at least publicly, though it has trimmed more deeply than in years past. Its executives have expressed their commitment to the debt business even as revenues from fixed-income, currencies and commodities have declined even more than some of its peers.
But commitment to a business doesn't necessarily mean commitment to the number of people in it. Behind the scenes, Goldman appears to be preparing quickly for a future with far fewer human traders than it has had in the past.
One example can be found in the "Goldman Sachs Algorithm," or GSA, a computer program that allows investors to trade U.S. corporate bonds without having to communicate with a human trader, according to a Financial Times article on Tuesday.
The idea is that Goldman circulates lists of bond prices on different trading systems, including Bloomberg, that can be acted upon immediately rather than just serve as a starting point with a human trader. The bank relies on computer programs to adjust the prices in real time according to market conditions and other trades, according to the Financial Times article. (Bloomberg LP is the parent company of Bloomberg News.)
Goldman is responding to a sea change in the $8.4 trillion U.S. corporate-bond market. While electronic trading hasn't yet proved to be the answer for this market, it's gaining an increasing share of the business, with the share of investment-grade notes traded electronically doubling in two years, according to a Greenwich Associates survey released in December.
The leading electronic corporate-bond trading platform, MarketAxess, has done fabulously well against this backdrop, with its shares surging 52 percent so far this year. But other startups have tried and failed to make their mark in the business, leaving it largely in the rapidly-becoming-anachronistric world of phone calls and emails.
Goldman's effort alone is unlikely to revolutionize the way investors trade corporate bonds, or even be a viable way for it to break into the electronic credit-trading business. A previous effort, GSessions, started fizzling out several years ago, with that network's creator, Chris White, departing the bank in 2015. And an algorithmic approach can only work if there's enough reliable real-time market data to process. This accounts for a tiny sliver of the corporate-bond market, where most bonds don't trade every day.
Moreover, Goldman's corporate-bond trading algorithm is intended for trades worth less than $1 million of highly rated debt, according to the Financial Times article. This isn't a business that's popular among highflying traders who like to take on risk and make big margins, especially as it becomes harder to eke out commissions with yields at record lows.
Still, it's a way for a bank to stick with its identity as a fixed-income shop in a low-yield, low-volatility world and straddle both worlds. It doesn't remove the bank from the equation, but it doesn't require as much expensive client-servicing either. And it gives the remaining human traders more time to focus on transactions that require more judgment and thought.
"It's pretty absurd if in this day and age you can only trade a bond through a human being," said Seth Merrin, the CEO of Liquidnet Holdings, which has a bond-trading platform.
Goldman may be right that the fixed-income market will once again prove immensely lucrative. But that doesn't mean it necessarily needs to pay big teams of people to keep the business running smoothly. It is taking some tentative steps to find out if it can have the best of both worlds.
This column does not necessarily reflect the opinion of Bloomberg LP and its owners.
To contact the editor responsible for this story:
Daniel Niemi at firstname.lastname@example.org