Treasury traders remain wary about adopting algos

Every day, decisions about when, where and how to trade billions of dollars of assets are delegated to client execution algorithms, which have become common features of the equities, futures and foreign exchange markets.

Yet in the $614 billion-a-day Treasuries market, the large dealers that dominate dealer-to-client trading have shown little interest in building these algos out.

Nevertheless, smaller players in the asset class believe algos could be the key to building market share.

A US rates trader at one large non-US bank believes the market “is ready for disruption in the way the FX market was a few years ago. Banks need to wake up to it.”

As is the case with interest rate swaps, there is a clear split in the Treasuries market between interdealer and dealer-to-client segments. In the latter, clients typically execute via requests for quotes, direct streams or voice trading—all of which involve executing via dealers. According to a 2020 report from the US Federal Reserve, the top 10 dealers at the time had nearly three quarters of volumes in the D2C market.

The electronic trading platforms cater more to dealers than to end-clients, and dealers don’t want to be disintermediated by algos

Hitesh Mittal, BestEx Research

Rates traders at the largest dealers have said that clients are not beating down their doors to ask for Treasury execution algos. A Coalition Greenwich survey from the end of 2021 reported that only around 25% of the 40 buy-side respondents were already using Treasury algos, and that none of these 11 respondents were expecting to use them more in the future.

Some believe the entrenched position of the large dealers is a key reason why algos have yet to take off in Treasuries.

“There is definitely a set of banks that are very well entrenched in the market, that have decent market share,” says the US rates trader. “There is going to be some reticence to change if you already have an established position in the market which gives you some kind of competitive edge.”

Special FX

The rapid electronification of FX and its numerous liquidity providers—most banks worth their salt have a foreign exchange desk—meant there were greater incentives within the asset class to create execution algos that clients could rent when they needed to execute large orders.

An FX execution algo will split up an order into smaller chunks to minimize the impact of the flow on the wider market. It will often seek to tap a bank’s internal flow before going out to interdealer and D2C venues and trading in the dealer’s name. The algo can act as an aggressor in the market and cross the bid/offer spread, or it can leave passive orders that take longer to complete but which enable the client to earn the spread.

FX algos increased in popularity as understanding grew about the impact large orders can have on the market. If a client sends a large order via request-for-quote to several dealers, the information leakage can mean the winning bank is unable to hedge the flow before the price moves against it; this, in turn, leads to worse prices when the client asks for another quote.

The fragmentation of liquidity in FX across multiple venues was another factor driving the increased take-up. Execution algos can knit together prices across numerous platforms, giving clients access to more sources of liquidity.

Similar dynamics exist in today’s Treasuries market. The dominant market position of BrokerTec, the former Icap subsidiary now owned by CME Group, is being challenged by the likes of Tradeweb, Fenics UST, LiquidityEdge and MarketAxess. Cboe has launched an unlit venue designed to trade large block orders in the interdealer Treasuries market. These venues all offer central limit order books or bilateral streams that could be accessed by algos.

However, Hitesh Mittal, chief executive of algo developer BestEx Research, claims large dealers have an “incentive not to” create algos in business lines where they are already powerful market-makers.

“The electronic trading platforms cater more to dealers than to end-clients, and dealers don’t want to be disintermediated by algos,” he says. “Would you rather make the bid/offer spread or charge commission? And the answer is always to take the spread.”

For this reason, algo innovation in the Treasuries market is coming not from the dealers with the biggest balance sheets, but from firms that are more willing to experiment with technology in a bid to gain market share.

UBS has offered Orca, an algo that takes top-of-book liquidity and tops it up with correlated bonds and futures. BNP Paribas is also looking to build in this space as part of its effort to combine its FX and rates business into a macro unit.

RBC Capital Markets is another firm that has pioneered the use of FX-style algos for large asset managers in the on-the-run Treasuries market. RBC’s algos focus on the most actively traded on-the-run Treasuries, which are more liquid and more electronically traded than their off-the-run equivalents.

“There are 300-odd Treasuries, but HFTs [high-frequency traders] only care about seven benchmark points because they match up perfectly with interest rate futures,” says Haider Ali, head of US Treasury trading at RBC Capital Markets in New York. “If a client wants to buy a very large amount of duration, and they want to do it over the course of 20 minutes in an intelligent way, we’ll let them source liquidity through our strategies that are very similar and that clients are very familiar with from an equities or FX landscape, like implementation shortfall.”

Ali explains that these strategies are designed to minimize slippage—either between the price when an order is received and when a broker implements an order, or from established benchmarks, such as time-weighted and volume-weighted average prices.

RBC has been offering a specific rates algo product since Q1 2018. It sources liquidity from the bank’s own internal inventory and from public central limit order books.

“We have a large number of institutional clients who use it and are pretty happy with it,” says Ali. He adds that as rates volatility has risen amid surging inflation and central bank policy changes, clients have become all the more cautious about the market impact of block trades.

“As volatility rises, as uncertainty increases, these kinds of tools become more useful,” he says. “We’ve seen a decent amount of pickup in the last two and a half years, really starting with Covid.”

Ali stresses that RBC still has a strong principal risk-taking business, which is there to handle more traditional approaches to risk management.

“We look at this as another offering which the market wanted,” he says. “But we are also there with our balance sheet and our capital to commit to the other 300-plus Cusips [an identification number for a North American security] that the clients are looking for, which tends to represent the majority of the outstanding volume.

“We’re able to warehouse risk where we need to where clients are asking us to commit risk, but we’re also happy to provide these tools to help them efficiently transfer risk in the market.”