Executive summary

Algorithmic trading, or electronic trading of financial securities based on algorithms, has seen a sharp spurt in recent years, thanks to advancements in cloud computing, quantum computing, Blockchain, artificial intelligence and machine learning.

Financial institutions make use of algorithms to route and execute trades at a speed and frequency impossible for a human trader to match. And the goalpost is being shifted continually as financial institutions look for ways to optimise data acquisition, modelling capabilities and underlying processes to speed up execution, improve market liquidity and generate higher risk-adjusted returns.

This growth in algorithmic trading, however, comes with its own set of risks. Regulators are increasingly concerned about the impact failed algorithms can have on the functioning and stability of financial markets. Malfunctioning algorithms can lead to significant losses, with the potential failure of financial institutions in extreme scenarios and systemic consequences for the economy. The possibility of algorithms being used for market and price manipulation is also of concern.

High-profile algorithmic trading incidents witnessed in recent years include the May 2010 US stock market flash crash and those of Knight Capital in August 2012 and Peet’s Coffee and Tea in September 2012.

Given the concerns, there has been increased regulatory scrutiny of firms using algorithmic trading strategies and systems through the Markets in Financial Instruments Directive (MiFID II)–RTS 6 and the Prudential Regulation Authority (PRA) Supervisory Statement 5/18. 

Model risk management standards and practices have also come under the regulatory scanner, with focus on the design, development, implementation and use of models. The Federal Reserve Board (FRB) and Office of the Comptroller of the Currency (OCC) SR-11-7 guidelines in the US has become the de facto regulatory standard for model risk management. Among other regulators, the Office of the Superintendent of Financial Institutions (OSFI) in Canada, too, has defined the E-23 guidelines covering expectations on enterprise-wide model risk management practices. Overall, there is an increasing expectation from regulators that trading algorithms should adhere to SR-11-7 principles.

In this paper, we provide an overview of the regulatory landscape for algorithmic trading, the key challenges faced by financial institutions, how the industry should respond and how CRISIL can support financial institutions in addressing the myriad of challenges.