Asia’s FX markets step to the algorithm

Buy-side use of automated trading programs builds, albeit from a low base

  • Banks are reporting a sharp growth in FX algo adoption among Asian clients, thanks to improving liquidity in some currency pairs.
  • The trend became more pronounced after the volatility seen in March and April due to Covid-19. Algos are said to have performed well during the crisis, partly dispelling fears that they work better in calm markets.
  • While G10 pairs still represent the majority of trades executed via algos in Asia, USD/CNH is gaining momentum, with the pair one of the top in terms of volumes in the region.
  • As firms gain more experience with automated techniques, they are open to trying more sophisticated algos to optimise execution – besides TWAP or VWAP.
  • However, liquidity during Asia hours in the marketplace in general is thinner compared with London trading hours, leaving firms with a dilemma over which timezone to trade in.

In the quirky and fragmented markets of Asia-Pacific, the use of algorithms to execute foreign exchange trades has struggled to gain a foothold. But banks in the region have seen a rise in algo adoption from buy-side clients in recent months, as more companies get to grips with the technology and recognise the potential cost savings available.

“The adoption of FX algo products has been gaining pace in Asia compared with around two years ago, particularly with deliverable and non-deliverable emerging market Asia pairs,” says Rob Ma, an executive director in fixed income, currencies and commodities execution for Asia at Goldman Sachs.

Algo usage is anecdotally lower in Asia’s FX markets than in Europe and the US, where around one-fifth of all participants use algos to trade FX, according to a 2019 survey by consultancy Greenwich Associates.

The region’s adoption lag has its origins in a cultural resistance to change among some buy-side firms, observers say. This has made the process of education more of an uphill struggle. Also, a lack of liquidity in some currency pairs in Asia has fuelled a perception – an unfair one, dealers argue – that algorithms aren’t effective in these markets.

But with trade sizes increasing in the region, banks have begun to convince clients that algorithmic execution can be cheaper in some instances. The coronavirus-related volatility that swept markets in March also provided a test for the nascent method. It was a test that, by and large, algorithms passed.

“During the disrupted markets that we experienced earlier this year where we saw extremely high volatility and extremely thin liquidity, clients turned to algo because sticking with the traditional static trading model no longer worked,” says Lisa Lee, head of e-sales for Asia ex-Japan at Crédit Agricole.

Dealers such as BNP Paribas, Deutsche Bank and JP Morgan are now seeing varying levels of growth in FX algo usage in Asia, with one banker describing the increase in adoption as “exponential”.

Algo trading won’t work for everyone, though. Not all clients will have the volume or size of trades to warrant using the technique. For these firms, it may still be cheaper to stick to the trusted method of risk transfer, where the client passes the whole trade to the dealer for execution at a fixed price.

“Trades by some of the counterparties in Asia might be smaller in nature, and that means a risk transfer is probably going to have a similar outcome to an algorithm,” says Vittorio Nuti, head of algorithmic trading for FX and listed derivatives at Deutsche Bank.

Clients also need to be aware that algo trading brings a change in risk. Users of algorithms take on the market risk of the trade, as an October report from the Bank for International Settlements points out. Trading via the risk transfer method means the dealer – and not the client – holds the market risk.

E-volution

The rise of algorithmic trading has been made possible by the electronification of markets. E-trading is common in equity and, to a lesser extent, fixed income. In foreign exchange, 57% of trades are executed electronically, according to BIS data from 2019.

Asia’s e-trading volumes are only a few percentage points behind the rest of the world, a 2018 report from trade body GFMA and consultancy KPMG suggests. This is primarily down to the large use of G10 currencies, which made up two-thirds of the flow.

But for local Asian deliverable currencies, for which Asian clients are responsible for 70% of volumes according to one liquidity provider, the report says electronic trading faces specific hurdles. For example the markets are fragmented; interbank trading of USD/onshore renminbi must happen on the China Foreign Exchange Trade System, while electronic USD/Korean won trades have to be done on two state-sponsored platforms. These trades are not allowed to flow through commercial electronic platforms.

The report also found that Asian market participants were less interested in multi-dealer platforms, and says secondary FX venues managed out of London and New York may not have the same focus on attracting Asian market participants.

A big growth area for our algorithmic execution over the last two years has been CNH, in conjunction with other emerging market currencies, such as IDR

Stephen Jani, JP Morgan

There are other reasons for the slow adoption of FX algos in Asia. Asif Razaq, global head of FX algorithmic execution at BNP Paribas, says mindset is one of them. The older generation of traders are traditional and used to relationship-based trading, he says. Some traders on the buy side saw algos as a threat to their jobs.

“That was quite a hurdle for people in that region. So there was a lot of pushback on the buy side. It took a while to change that mindset,” says Razaq.

Many firms also have used aggregators to pull in prices from a range of different sources, and felt that would be the best way to gather liquidity. Nuti at Deutsche Bank says the cost of running these aggregators is for many clients higher than using FX algos.

When you add up those costs versus the relatively light volumes that the clients are doing, trading algorithmically can get expensive, he says.

Lee at Crédit Agricole agrees there hasn’t been sufficient pull from clients to rapidly take up algo trading. She points to the misperception among some clients that algos are only suitable for super liquid G10 currencies, and are not effective for less liquid Asian pairs.

Many Asian clients also tend to transact in small trade sizes. This means the cost of trading via risk transfer was not dissimilar to doing so via an algo. Without a clear cost advantage, algos have struggled to gain traction among buy-side firms.  

Automatic for the people

In recent years, though, Asia dealers have noticed a pick-up in algo activity. There is a spectrum of growth across banks. For BNP Paribas, FX algo adoption in Asia has grown threefold since January 2018. At Crédit Agricole, the proportion of Asian clients trading algos has more than doubled from a figure of around 5% in 2018.

“Algo volume in Asia is still a minor portion of the overall FX volume but it’s growing exponentially,” says Crédit Agricole’s Lee.

JP Morgan says it has seen in excess of 80% growth in FX algo usage across the region. “We certainly didn’t see early adoption in Asia, but over the last two years we are the fastest region of growth for algorithmic execution,” says Stephen Jani, the bank’s Asia-Pacific head of e-sales for fixed income, currencies and commodities.

About 20% of JP Morgan’s clients that use FX algorithms are based in Asia, Jani adds.

For Deutsche Bank, FX algo usage in Asia increased by about 10% in the past year, and the bank expects the adoption to speed up in the next few years.

dollar-yuan
US dollar/offshore renminbi has been one of the most actively traded currency pairs on FX algos in Asia

The growth has come from a few areas. First is education. Jani at JP Morgan says buy-side firms are becoming more familiar with the space, helping drive penetration. Similarly, BNP Paribas has shown clients transaction cost analysis that suggests algos can outperform trading via risk transfer.

“We are able to show that if you want to trade 100 million against risk pricing from a bank, the algo typically would have a high percentage chance of outperforming the institutions. And when you put that quantitative study in front of somebody, that sets up experiments with the use of algorithms, and they start getting comfortable with it. It made the transition much easier for them,” Razaq says.

Lee at Crédit Agricole says that as the size of FX trades has grown in the region, so has the attractiveness of doing them for less money via algos. Currently, the average order size in the US and Europe is larger than in Asia, she says. “But I’m confident that in time, algos will be adopted by Asian clients in a similar way,” she adds.

A test of the market came during March, when the onset of the coronavirus pandemic sparked runaway volatility across asset classes. The growth of algos during the previous 10 years had come at a time of prolonged low volatility. The common view was that clients felt more comfortable taking on the market risk of trades when the chance of slippage or losses was low.

However, algos performed well during the stressed market conditions, contrary to predictions. Ma at Goldman Sachs says algo execution volume in the region during March was two to three times higher than in the same period last year.

Current flows

In deliverable currencies, the bulk of algo usage in Asia is for liquid G10 currencies, notably Australian dollar and yen, Jani at JP Morgan says. However, local currencies are gaining momentum, particularly USD/offshore renminbi.

“A big growth area for our algorithmic execution over the last two years has been CNH, in conjunction with other emerging market currencies, such as IDR,” Jani says.

Goldman Sachs says USD/CNH is one of the top currency pairs for its algo in terms of volumes in the region.

“Half of algo volumes traded by Asia clients are still G10, with CNH being one of the most active currencies in Asian clients’ algo execution,” says Ma. “USD/CNH is still seen as an emerging markets Asia currency pair but it’s among the top three in terms of volumes in the past few years.”

Elsewhere, non-deliverable forwards are a growth market for algo execution, as regional clients push banks to find cheaper ways to trade the products via algos. Globally, algo trading of NDFs is still “nascent”, the BIS notes in its October report.

When it launched its NDF suite earlier this year, Barclays said the demand had mostly come from its Asian client base with exposures in the local currencies. Popular NDF currency pairs among Asia clients include USD/KRW and USD/INR, Goldman’s Ma says.

As clients gain more knowledge of FX algo products and the market microstructures, they are leveraging the more sophisticated algos to optimise their execution

Rob Ma, Goldman Sachs

Dealers say passive algos are more popular with clients in Asia than other regions, while limit-based strategies are also used.

Lee says clients tend to dip their toes into the algo market by taking a small order using so-called passive algos that post bids and offers on venues and wait to trade at the client’s required price. As they gain a greater understanding of the mechanics of algo trading and transaction cost analysis, they start to trade more frequently and in bigger lots.

Clients that were earlier adopters of algos tended to have experience in other asset classes where algos have been more entrenched, such as equities. That meant they were initially more comfortable with algo styles such as those based on time weighted average pricing (TWAP) or volume weighted average pricing (VWAP), which are prevalent in equities. These techniques split up large trades into smaller chunks and execute them at a targeted price, determined as the weighted average price for that trade. They are designed to have less market impact than executing a single large trade in one go.

“In the past when clients were not as well educated on execution algos in FX, they were more inclined to pick an algo they were most familiar with. An example would be if they had an equity trading background, a TWAP or VWAP algo might be the algo of choice,” says Goldman’s Ma. “But over time as they gain more knowledge of FX algo products and the market microstructures, clients are leveraging the more sophisticated algos to optimise their execution.”

Time to trade

A practical question for FX algo clients in Asia is which timezones to use for executing their orders, as currency pairs can have different liquidity profiles during the day. While most G10 pairs have the greatest liquidity in London hours, pairs such as USD/CNH, USD/SGD and USD/JPY can have similar liquidity conditions in Asia and London timezones.

For other Asian pairs, liquidity becomes progressively worse to trade outside the local timezone, says Vittorio Nuti at Deutsche Bank.

There are no hard and fast rules on how clients trade, however. Clients in Asia can start algorithmic execution in Asia trading hours and finish before London hours kick in. They can also wait until London checks in, at around 2pm–3pm Asia time.

Larger clients such as global asset managers and global hedge funds generally have execution desks across regions. This enables them to roll trades from one centre to another, depending on liquidity in certain timezones or the requirements of the individual trade.

“Some clients will want to get the risk done as soon as possible, others might want to execute at the optimal time of day based on the historical intraday liquidity profile,” says Rob Hutchins, head of execution services for fixed income, currencies and commodities in Asia at Goldman Sachs. “Therefore they may choose to wait until a different timezone in order to target the best liquidity, with the tradeoff of assuming more market risk for the intervening period.”

Goldman’s Ma says that during March’s volatility, a number of clients who historically traded on benchmarks during London hours started to adopt algo execution for the first time during Asia hours to minimise the market risk from the heightened volatility.

Editing by Alex Krohn

  • LinkedIn  
  • Save this article
  • Print this page  

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact customer services - www.fx-markets.com/static/contact-us, or view our subscription options here: https://subscriptions.fx-markets.com/subscribe

You are currently unable to copy this content. Please contact [email protected] to find out more.

You need to sign in to use this feature. If you don’t have a FX Markets account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an indvidual account here: