Better tech brings threat of two-speed trading in fixed income

When expensive new technology comes along, it’s usually the bigger, richer firms that can benefit from it the most.

And so it goes with automated trading in fixed income, a practice that has so far lagged its equity and foreign exchange cousins. But as technology and access to data continue to improve in the asset class, some buy-side traders believe the costs involved in building the most up-to-date systems will invariably lead to a divergence in performance between the algo ‘haves’ and ‘have nots’.

“Unless you have people who are able to do this internally, it’s very difficult to justify the investment,” says Eric Boess, global head of trading at Allianz Global Investors. “For example, the number of firms that are big enough to support a pure data science team in trading, versus sharing the data scientist team with portfolio management is limited.”

The technology in fixed income isn’t yet advanced enough for both-hands-off-the-wheel algorithmic trading, as is the case in cash equities or spot foreign exchange. But asset managers are using automation in parts of the trade lifecycle to give them an edge over peers and increase efficiency. For example, firms are using the data gleaned from automated transactions to compare the accuracy of counterparties’ pricing. Or they are using automation to save money by channeling more trades through the desk.

Large asset managers are now automating anywhere from 50% to 90% of government bond trades, firms report. Figures are lower for corporate bonds.

But a class divide is looming. Only the firms with the deepest pockets can afford to buy the kind of data needed to power the latest automated trading engines. No data, no dice.

“It’s very hard to get good aggregated and normalized datasets you can rely on. Data is quite expensive still, and even if you buy it, it takes quite some time to normalize it and understand how you can utilize it since the data is far from perfect,” says Kasper Folke, co-head of global trading at Nordea Asset Management.

Once you acquire the data, you then need to interpret it—which requires human capital. Again, firms with bigger desks have an advantage as they are more likely to have the specialists for the job.

“It’s not a matter of pure payments, it’s a matter of people dedicated to such projects. On the trading side, they need to spend time after the markets close to analyze those evolutions and put these projects in place. It’s an indirect cost and requires time,” says Vincenzo Barbagallo, head of trading at Generali Insurance Asset Management.

Once the tech frontrunners pull ahead from the chasing pack, the gap could continue to grow because access to the latest gadgetry helps the largest managers evaluate its worth and justify continued investment, observers say. Conversely, the problem for the tech-impoverished also-rans is they don’t know what they’re missing out on.

“The more technology and data you can have, the better you become in assessing the value. A big challenge is that when you don’t have it, you don’t have the measurability to understand how much it can help you,” says Folke.

Automatic for the people

Automation in fixed income has been on the rise for some time, and buy-siders are at different stages of adoption.

Allianz’s Boess says the industry’s current capabilities fall well short of full algorithmic execution. Instead, the tools available to the buy side reflect the need to improve productivity by using decision-tree style execution where a computer follows simple execution directions but does not respond dynamically to market conditions. The firm has looked to take the process a step further and apply machine learning technology to areas such as price history and dealer axes so that a computer can trade in a more “human” way for illiquid or large-sized orders.

Inés de Trémiolles, global head of trading at BNP Paribas Asset Management, says the manager has seen its trade count go up by 50% in the last five years, an increase that would have forced the desk to double its headcount if it had retained its 2019 technology, she estimates. Instead, they’ve been able to keep staff numbers steady.

The firm now automates—without any trader intervention—roughly nine of every 10 government bond transactions, which de Trémiolles says allows them to quickly trade close to the mid-price.

“When a portfolio manager sends an order for a government bond, they want that rate that they’re seeing on the screen. Our role is to go as fast as we can to execute without moving the market so that we reduce that time and market slippage for the portfolio manager,” she says.

Thanks to more available data in fixed income, people are throwing technology that has been used for years already in other asset classes at fixed income and see what they can make out of it

Eric Boess, Allianz Global Investors

Norges Bank Investment Management runs roughly half its tickets through an automated tool. The firm first applied automation to US Treasuries trading several years ago, with European government bonds following more recently, says John Sullivan, global co-head of fixed income trading at the firm. Credit trading remains overwhelming voice, though.

Asset manager Generali also executes over half of its government bond trades, on average, and roughly a quarter of corporate bond trades in an automated fashion, which sees the trading desk directing orders to automated tools that select counterparties and execute trades but still allows traders to intervene if needed. Barbagallo says he expects those figures to increase in coming years.

After automating much of its rates trading, Nordea IM’s Folke says the manager can now compare the competitiveness of their counterparties and how well their indicated pricing levels match the prices at which they trade in different market conditions.

Ways to go

But the lack of adequate data and need for continued technological development means true algorithmic fixed-income trading remains a thing of the future. Automation in fixed-income markets is limited to the most liquid government bonds, some interest rate swaps, and small-sized trades in corporate bonds. And even then, most trading happens via request-for-quote and so systems are effectively automating only the request and execution process.

Some fixed income markets have closed the gap in recent years, notably those for government bonds and US investment-grade credit, as users migrate to platforms. Nearly half of investment-grade bond trades on the official US reporting platform Trace passed through Tradeweb and MarketAxess in October, according to aggregation of each firm’s volume market share estimates. The figures are up from just over a third of volumes at the start of 2020, although growth has plateaued more recently. Platform volumes for high-yield bonds are lower but have nearly doubled in market share over that time.

But electronification of corporate bond trading has lagged efforts in markets like spot FX and cash equities, where prices stream on multiple venues and buy-side firms use automated execution strategies.

However, these asset classes can act as a blueprint for advances in fixed income, experts say—particularly if traders are able to access more, and better, data.

“One of the biggest developments over the last couple of years, thanks to more available data in fixed income, is people are throwing technology that has been used for years already in other asset classes at fixed income and see what they can make out of it,” says Boess.

The next step will be to integrate platform offerings with streams from liquidity providers, centralizing pricing, axe information and available liquidity on order and execution management systems. That would allow algos to monitor and automatically choose the best price to execute on across different liquidity pools.

Folke at Nordea AM expects that the benefit of these integrations will become apparent in the coming years as banks, venues, and technology providers improve offerings that will allow buy-side traders to view fixed-income trading opportunities on a single screen without having to switch between platform and vendor pages. And it’s at that point that the differentiation between the buy-siders with teams of data scientists and technologists and those without may become clear.

The model faces challenges, though, with platforms reluctant to cut into existing revenue streams and technology providers walking a fine line to avoid regulation as multilateral trading facilities.

The next stage to properly move fixed income automation forward depends, in part, on the ability for firms to access quality pre- and post-trade information. Traders complain that the pricing they receive from liquidity providers can be stale when they want to trade on it, and trades that occur off exchange or are subject to lengthy reporting deferrals limit market transparency.

“Automation and fixed income still has a lot of catch-up to do versus other asset categories, such as equities and FX. Trading automation needs data input, and in fixed income we expect availability and quality of data to grow in the years ahead,” says Jan Mark van Mill, managing director, multi-asset at APG Asset Management.

Automation also requires reliable pricing inputs so algorithms can make accurate predictions, direct trades to the correct sources, and determine a fair price. Sullivan at Norges Bank Investment Management says the buy side has to trust these inputs.

“When you set parameters, you say you’re willing to trade within plus or minus this [price] point. If you’re not comfortable with that baseline to begin with, then it takes away the point of streamlining the process,” he says.

Plug and play

Trading platforms are playing their part in increasing automation. Tradeweb’s AiEX tool, for example, allows users to send orders to the venue via API, with trades of a certain size then sent automatically to dealers for quotes. The dealers are selected based on pre-set criteria, and the trades executed if the best price is at a target level.

While direct APIs between the venue and execution management systems have been used by a handful of asset managers and hedge funds, electronic trading experts say demand has been low, given the investments required.

Even though fixed income markets are not yet ready for full automation, Cathy Gibson, global head of trading at asset manager Ninety One, says smaller buy-side firms can keep pace for now by using platform-owned automation tools.

“The technology is reasonably inexpensive, so that’s not a barrier to entry. Because it’s provided by platforms, even if you don’t have an order management system, it wouldn’t be as efficient, but you could still access low-touch trading strategies,” says Gibson.

Still, applying the technology to existing trading patterns can take time, even if platforms and software vendors take care of the technical work.

Also, even when fixed income markets are able to become more truly automated, Nordea’s Folke says if smaller firms have to make do with the existing platform-owned workflows they won’t have the same flexibility to compare prices across markets as the bigger players that have made the requisite investments.

But van Mill at APG says the benefits from investments made by large managers will eventually filter their way through the market.

“The technology finds its way into the smaller asset managers with a bit of time lag. If you look, for instance, at what an average asset manager can get out of an equity execution management system, there’s no first and second tiers,” he says.