Broker-dealers, proprietary trading firms and exchange operators have more closely monitored the equity trading within their four walls for a number of years, as per regulatory mandates and their own strong desire to stay out of the headlines that bedeviled the industry during the global financial crisis of 2008-2009.
This trend has accelerated amid heightened volatility and the shift to working from home that began in March 2020 due to Covid-19. With employees working and trading remotely, the pandemic underscored the importance of deploying technology to perform equities market surveillance more efficiently and in a smarter way.
Regulatory surveillance is mission-critical for the Financial Industry Regulatory Authority (FINRA), according to Stephanie Dumont, Executive Vice President, Market Regulation and Transparency Services at the self-regulatory organization.
“Of the priorities day-to-day that we’re really focused on in our department, our regulatory surveillance program is number one,” Dumont said Oct. 7 at the Security Traders Association market structure conference in Washington, DC. “We are constantly thinking about ways to be better, improve, and innovate in our relationships with our regulatory partners, our SRO exchanges, and being able to partner and work together on surveillance and cross-market regulatory initiatives.”
US regulators have brought on a number of enforcement cases over the last few years, while in the UK, the Financial Conduct Authority published guidance last year stating that firms were expected to review and update their risk assessments and change their surveillance systems to detect any new or heightened market abuse risks. Julia Hoggett, Director, Market Oversight at FCA, said in a 2020 speech that the usual levels of recording and surveillance were not possible early in the pandemic, but firms overcame these challenges.
“Our expectation is that going forward, office and working from home arrangements should be equivalent – this is not a market for information that we wish to see be arbitraged,” Hoggett said. “New communication mechanisms, before they are used, should have controls in place where required and their use be approved by firm management. The regulatory obligations have not changed, the how may be changing, but the what remains the same.”
Joe Schifano, global head of regulatory affairs at Eventus Systems, highlighted that the proliferation of communication channels, mobile devices and encrypted messaging services such as WhatsApp has made it easier to share information covertly, making it more critical for firms to employ state-of-the-art surveillance technology and procedures. Firms now have access to vast amounts of data, inexpensive computing power, and innovative technologies to help develop automated compliance and risk-management tools.
FINRA said in a paper last year that some firms are developing surveillance and conduct monitoring tools using deep learning models which are built on an artificial neural network. Algorithms can process large amounts of underlying data from disparate sources and in different formats, such as text, voice and video, in a similar manner to how neural networks function in the brain.
From Rule-Based to Risk-Based
Artificial intelligence provides the ability to capture and surveil large amounts of structured and unstructured data in order to identify patterns and anomalies. FINRA said: “Firms indicate that these tools offer the ability to move beyond traditional rule-based systems to a predictive, risk-based surveillance model that identifies and exploits patterns in data to inform decision-making.”
Schifano continued that the unlimited potential of automation and machine learning is used in Eventus’ Validus platform. “Validus analyzes and identifies the most important alerts by asking the kinds of questions a human would ask,” he said.
Legacy systems have traditionally set granular definitions of possible suspicious behavior which generates many false positives. For example, if one trading desk sends a buy order around the same time as another is selling the same security, the system may flag a suspicious wash trade, even if the transactions were not problematic.
“Defining a very specific behavior such as a wash sale and producing an alert is not smart enough anymore,” said Schifano. “Technology has greatly advanced the ability to surveil larger datasets in a smarter way.”
The number of false positives multiplies as more data is reviewed and as activity increases in volatile markets, and firms may try to solve the problem by cutting down on the amount of data reviewed. Validus’ approach is to take in all the transaction data, regardless of predefined parameters and perform reviews on a case-by-case basis using parameters such as how often the behavior occurs, market share, how many trading desks are running into each other, and whether the behavior began recently.
“Through various types of automation – machine learning, statistical analysis, trend spotting – we can answer these questions accurately and efficiently, helping our clients gain a clear view of all activity occurring on their watch,” Schifano added.
In addition, the platform creates a robust audit trail documenting why alerts may not be classified as potential manipulation after they have been investigated, which can be reviewed by regulators if necessary.
Effective surveillance is very much related to the availability of data. Schifano said data lakes have improved in the US because of the CAT process and that data is being used in more innovative ways. The CAT or Consolidated Audit Trail was introduced in 2018 to aggregate trade for US equity and options markets following the “flash crash” in 2010 when shares suddenly fell in value.
However, firms need to ensure they keep pace with ever growing amounts of data and the increased speed of technological change. In addition, they want holistic surveillance where they can get an overall picture across interoperable systems and asset classes, including new digital assets.
“The future of market surveillance is very much tied to a firm’s ability to maintain the pace of technological change,” said Schifano. “They have to be able to handle the data, look for anomalies and have a system that is flexible enough to easily change.”
