Traders are turning to AI to manage market impact on large orders, finds ESMA paper

EU watchdog has released an article exploring the use of artificial intelligence and machine learning in the markets, noting traders are utilising technology to minimise impact.

Artificial intelligence (AI) models are increasingly being used by traders, brokers and financial institutions to optimise trade execution and post-trade processes, resulting in reduced settlement failures and market impact of large orders, an article by ESMA has found.

In its latest article, the EU watchdog details the use cases of AI in EU markets, alongside examining the technology’s development and potential risks associated with its adoption.

ESMA noted that AI is most promisingly applied in the execution phase of the trading life cycle, particularly when brokers attempt to minimise costs stemming from market impact caused by filing an order.

“Accurately estimating market impact has become particularly important for investment banks and other brokers operating low-margin businesses,” said ESMA. “Nevertheless, this quantity is notoriously hard to model, especially for less liquid securities, for which data on comparable past trades is scarce.”

Machine learning (ML), in this instance, allows brokers and institutional investors to minimise market impact of large orders by determining how they can be split amongst venues and trading periods optimally to minimise market impact and transaction costs.

The EU regulator also highlighted that AI can underlie specific trade execution algorithms that optimise the costs involved in the execution of a trade that has already been placed, by minimising its market impact, especially in the case of large orders.

Elsewhere, ESMA noted that AI models were being used by institutional investors to enable more efficient post-trade processing, particularly by optimising the allocation of liquidity in the settlement cycle.

In the pre-trade phase, ESMA found that AI models can be leveraged by investors to analyse signals in asset prices and identify investment opportunities, which can either be evaluated by a human decision-maker or be part of algorithmic trading strategies.

The quality of datasets used in the learning phase was noted as a widespread concern for the material impact it could have on the outcomes and performance of AI and ML applications. To take advantage of AI, data has been stressed by industry experts as a necessary condition for successful technology.

“AI depends on data as its ‘fuel’: the success of AI tools is highly dependent on data quality and poor-quality, noisy data can easily result in unreliable models,” highlighted ESMA.

ESMA noted that the adoption of AI is not likely to lead to a fast and disruptive overhaul of business processes, with technological constraints, client’s preferences and regulatory uncertainty playing a role.

“[Market participants] welcomed the prospect of a clear framework for the effective and trustworthy use of AI to help decrease the wariness that many still have towards its adoption,” added ESMA.