Credit robots take over junk bond markets

New York

CREDIT-trading robots, having already taken over much of investment-grade bond pricing and trading, are getting sophisticated enough to move into less-liquid corners of debt markets.

Bank of America executed 2.7 times more trading volume globally through its credit bots in the first half of 2021 than in the first half of 2020. Much of that recent growth comes from areas such as junk bonds and emerging markets, said Sonali Theisen, the bank's head of fixed income e-trading and market structure.

JPMorgan Chase's algorithms now trade several thousand different bonds in the US and the Europe, Middle East and Africa region, up from around 250 at the beginning of 2020, the bank said. At Barclays, algorithms now price more than half of quote requests globally.

The pandemic has sharpened these algorithms, as big market swings helped developers better understand how securities perform in stressed markets. Now, price movements have settled down - making it easier for automated trading to get involved in transactions in areas including emerging markets.

"The Covid crisis gave the industry a whole set of new data on how markets react in a crisis," said Kevin McPartland, head of research for market structure at Coalition Greenwich. "That's very helpful for the folks writing the algorithms, to understand what might happen in the worst case."

The greater automation is allowing banks to process more trades and enabling human traders to focus on larger, more complicated transactions. Trading volume in the US corporate bond market has grown 39 per cent over the last five years as companies have gone on a borrowing binge, while most banks have kept the number of traders they employ stable or even cut headcount a bit.

"One of the founding principles when we went and asked the bank to invest in algos was, 'We think we can make our people 10-20 per cent more productive by building this technology'," said Drew Mogavero, deputy head of credit products at Barclays. "When you're not having to stare at a request-for-quotes screen and just punch numbers, you have more time to do those higher value-added things."

Most algo trades now are responses to requests for quotes on electronic credit trading platforms such as those operated by Tradeweb Markets, MarketAxess Holdings and Trumid Financial. The platforms have seen explosive growth of automated trading, especially in high yield.

The share of automated trades for US high yield at Tradeweb was 38.3 per cent in the third quarter, up from just 5.1 per cent 3 years ago, said a company spokesperson. Algo trade responses on MarketAxess' platform were up 17 per cent in the third quarter compared to the same period in 2020.

"The algos are getting better," said Richard Schiffman, head of open trading at MarketAxess, where banks can trade with clients and other parties. "It's a steadily improving science that all these firms are performing."

For years, sell-side trading desks used algorithms to price and execute simple trades such as transactions involving smaller-sized US investment-grade bonds or heavily traded high-yield securities. Now banks can link their various credit businesses and figure out the best use for any bond that they end up buying, such as whether to package it into an exchange-traded fund or find a client to sell it to, based on risk tolerance and other factors.

And the growth of exchange-traded funds and portfolio trading is providing more buyers, more data points, and more opportunities for automation.

One reason algos are increasingly needed to assist human traders: The growing size of the corporate bond markets. The total outstanding value of bonds in just the Bloomberg US Investment Grade Corporate Bond Index has grown from US$4.3 trillion in 2015 to US$6.2 trillion now.

"People sometimes think of the relationship between algorithmic and human market-makers as a zero-sum game, but we never have because credit markets continue to grow," said Sanjay Jhamna, head of EMEA credit at JPMorgan. "The growing volume of tickets and number of issuers mean added pressure on traders to respond to more inquiries."

Improving tech is allowing dealers to respond to a much higher rate of quote requests. Before algo development, many electronic quote requests for small traders or odd sizes were ignored by dealers, with human traders focused on executing more-profitable big trades. At one large sell-side credit trading desk, as many as three-quarters of electronic quote requests went unanswered before the adoption of algo trading. Thanks to algos, that rate has been flipped, and the bank is responding to about three-quarters of tickets, said a person with knowledge of the matter who asked not to be identified because the information is private.

Emerging markets are a major challenge for credit-trading bots now. Fragmented markets and a lack of transparency and data make developing markets much more difficult to trade via algos. But technology is developing fast.

"EM is coming into focus for greater efficiencies, and I think algos will start to play a bigger role," said BofA's Theisen. "I think of EM as more in the toddler phase versus US investment grade, where the algos have grown into adolescence."

Morgan Stanley is actively improving the capabilities of its high-yield algo and extending its tech into emerging markets credit, said Maryanne Richter, the bank's head of credit electronic trading strategy.

"We are taking the underlying framework of the corporate bond algo and extending it to the different markets which are much less transparent, much less liquid, and much more idiosyncratic," Richter said. BLOOMBERG