Algos make market thinner, but not less liquid – BIS

Trading robots change market microstructure, central banks conclude; new liquidity metrics may be needed

The growing popularity of execution algorithms is resulting in thinner foreign exchange markets – but is not hurting liquidity – according to a new study from the Bank for International Settlements. New ways of measuring liquidity may be needed as a result.

Published today (October 30), the study is the work of participants from 22 central banks and draws on a survey of 70 market participants. Although it describes the growth of algos as a “net positive” for FX, it also identifies new risks – among them is a change to liquidity dynamics and the undermining of common liquidity metrics.

“Traditional order book-based indicators may no longer be a good proxy to assess liquidity conditions,” the study notes.

That’s because algorithms operate by cutting an order into slices the market can easily digest, and – when operating in passive mode – will place those orders close to the best bid or offer at that point in time. As each transaction occurs, a new order will be placed – again, close to the best prevailing price. In contrast, traders operating manually will place resting orders at various levels of the order book.

So, a market with high levels of algo participation will appear to be shallower than one that is populated by human traders – a phenomenon described to the central banks during their research.

“Several market participants mentioned in interviews conducted by the study group that visible depth in public order books is lower today than it used to be. While this phenomenon is largely driven by the automation of FX trading, there are signs that [execution algorithms] specifically are contributing to the thinning of the order book as they slice large orders into smaller ones and spread them over time,” says the report.

Traditional order book-based indicators may no longer be a good proxy to assess liquidity conditions

Bank for International Settlements

A thinner order book is generally seen as a less-liquid market – one that has less capacity to absorb trades without the price moving. This may now be an outdated view.

“As long as the order book is replenished fast enough, a thinner order book does not necessarily reduce market functioning. As velocity of trading increases with electronification, some market participants point out that liquidity replenishment, ie the rate at which new top-of-book liquidity is renewed, is what matters for a well-functioning market,” the study adds.

As a result, market participants – including central banks who want to monitor market function – may need “novel liquidity indicators” that track the rate of replenishment, rather than market depth. Those new metrics may be costly and require new technologies, the central banks warn.

Execution algorithms currently handle up to 20% of volume in spot FX, according to the study – with customers ranging from large asset managers to corporates and central banks.

The dark side of internalisation

Overall, the central banks find algos offer a number of benefits to the FX markets – and that they coped well when volatility surged in March.

The study uses data from vendor BestX to compare the performance of three broad families of algorithm against the cost of executing with a dealer. It found that all three algo families beat the dealer prices on offer (see figure 1).

Other benefits are mentioned, too: by quickly reacting to new market information and providing updated prices in a timely fashion across a plethora of trading venues, algos can improve the matching process in a highly fragmented market, the study argues. 

But there are downsides. Algos transfer execution risk from the dealer to the client, and it can be difficult for a user to know whether a specific algo “does in fact deliver on its promise of superior execution quality,” the study notes.

A more-distant concern the central banks raise is the indirect impact that growing algo usage could have on the market. A portion of algo flow will typically be matched off against the dealer’s other desks and clients – or internalised – before going out to the market. This is typically presented as a benefit for the client, because it reduces market impact, but the study notes a possible downside. If too great a share of trading is executed away from ‘lit’ venues, then it will become increasingly difficult for participants to calibrate their pricing models.

“From a market functioning perspective, a trade-off can emerge between price discovery and internalisation. What is optimal for an individual market participant may not be optimal for broader market functioning,” the report states. “The individual incentives for dark trading to limit market impact contrast with the market-wide desirability of lit trading to ensure a smooth price discovery process.”

While the report notes that current levels of internalisation “are not seen as negatively impacting the price discovery process at the moment”, the Bank for International Settlements invites market participants to pay closer attention to the phenomenon over time. It also underscores the need for central banks to deepen their understanding of this area of analysis as it could be relevant for other markets undergoing rapid electronification, such as fixed income.