Winning the Compliance Battle on Multiple Fronts: Tackling Market Abuse and False Positives
With the emergence of more regulations – including Market Abuse Regulation (MAR), MiFID II and the SEC’s Regulation Best Interest, financial services organizations (FSOs) around the world are continually redrawing the battle lines in their quest for compliance. And these highly prescriptive regulatory directives are clashing with exploding trading volumes, making the struggle all the more real. Analytical models have become fundamental tools in the FSO compliance arsenal. By one estimate firms deploy an average of 80 models – but this, in turn, has led to other problems. High volumes of false alerts are inundating compliance analysts.
Let’s attach some numbers to the problem. Options trading volumes have grown over 600% in the last 15 years. For a firm with 200,000 daily transactions (and 80+ typologies), this could result in over 700 alerts generated in a single day. More significantly, the vast majority of these are false positives. PwC’s 2019 Market Abuse survey revealed that of approximately 40 M trade and communication alerts generated across the 17 financial institutions surveyed, only 1 in 12,000 alerts resulted in a STOR filing, indicating an actual false positive rate of 99.99%.
But these sheer numbers mask the true problem – which is that market abuse is rarely, on the surface anyway, black and white. Furthermore, bad actors are constantly adjusting their strategies, making trade surveillance a constant cat and mouse game.
The fact is – detecting market manipulation can be incredibly complex. Consider Insider Trading for example. Accurately identifying instances of insider trading relies on a firm’s ability to detect and understand what actions a trader took (e.g. buying/selling instruments), and when those actions were taken, relative to a specific material non-public market event that resulted in a significant upward (or downward) price swing.
While this definition of insider trading seems fairly straightforward, in reality, accurately identifying it is much harder. Here’s why:
- Differentiating between material and non-material events. Identifying insider trading means you must be able to first distinguish between material and non-material market events, as identified through various news sources. To be considered "material" the market event must be significant enough to change the company’s stock price. So, for example, getting an early copy of a quarterly report that reveals a pharma company is moving to a new location probably wouldn’t be material, but getting a heads up that that same company is going into early trials with a COVID-19 vaccine would be an entirely different situation altogether. If your surveillance technology can’t distinguish between the two, you won’t get off first base.
- Traders find creative ways to avert detection. Traders can spread their activity over several, related instruments to try to fly under the radar and disguise their behavior. For example, beyond trading stock, they can buy debt securities (bonds), or trade on futures or options contracts. When traders spread buying and selling across different instruments, insider trading can be harder to detect.
- Static thresholds don’t adapt to dynamic market conditions. Insider trading is also tied to large price swings and for this reason most surveillance systems employ thresholds to detect these changes. The problem with this approach is that thresholds, by definition, are static. In contrast, capital markets, by nature, are far more dynamic, as the current pandemic has demonstrated. Daily price swings of three percent – which might be considered excessive under ordinary circumstances – would be low in today’s market, which has experienced extreme volatility due to COVID-19. The reality is – static thresholds don’t work in highly dynamic markets. Set the threshold too high and you risk missing something. Set it too low and you’re deluged by false positives.
- The right window – finding that perfect balance. To identify potential insider trading, a surveillance system must look back at the trades executed by a regulated employee preceding the material market event. But what’s the magic time frame? Look back over too short a period of time and you could risk missing something important. Cast a much wider net and you could end up generating excessive false alerts.
- Why analyzing profit/loss isn’t foolproof. Some firms believe that adding in and analyzing additional factors, like the profit a trader is making, can be a surefire way to reduce false alerts related to insider trading, because you can weed out less profitable transactions. But while analyzing profit and loss can be helpful in terms of detecting market abuse, it’s not a foolproof measure. For example, there have been known cases of attempted insider trading where the trader didn’t generate a profit (due to the timing of the trades or misinterpretation of information). But regardless of the end result, the data and communications revealed that the intent to commit insider trading was still there.
If these challenges sound familiar to you, holistic surveillance may be the answer. Here are five key ways that NICE Actimize’s SURVEIL-X Holistic Surveillance solution can help your firm reduce false positives and stamp out market abuse, including insider trading.
- Applying NLP to distinguish between material and non-material events. SURVEIL-X can ingest data from over 19,000 news sources, including Dow Jones. But it doesn’t just ingest news, it uses advanced NLP (Natural Language Processing) analytics and entity extraction techniques to identify the news’s relevance, sentiment and potential market impact. NLP cuts through the clutter to understand the context of the news, and based on this, assigns the news a relevancy score. The higher the score, the greater the likelihood it represents a material event.
- Cross-product analytics nab market abuse that could otherwise fall through the cracks. Using advanced cross-product analytics, SURVEIL-X can also detect insider trading across the entire capital structure of a firm, going far beyond the company’s listed stock. For example, this can include derivatives as well as any debt that the company may have issued.
- Finding a way around market volatility. As noted above, surveillance solutions that use fixed thresholds aren’t adaptable to volatile market conditions. One way around this is to deploy a solution that can automatically adjust alerting based on market volatility. For example, SURVEIL-X continuously compares fluctuations in individual stocks to fluctuations in the market index as a whole, and (at a firm’s discretion) can use this real-time information to suppress alerts related to price spikes that are not line with (significantly lower) than how the market is trending. This helps firms find a way around high levels of false alerts due to market volatility. Some analytical models can even fine-tune themselves by automatically adjusting for market activity.
- Anomaly detection offers a better way. One way of addressing the question of ‘look back’ time windows is to simply not rely on them alone. For example, using its advanced anomaly detection, SURVEIL-X applies unsupervised machine learning to identity changes in behaviors and flag outliers instead. These outliers might include communications or trading activity that’s atypical for the individual trader. Rather than looking for a specific instance of suspicious behavior (e.g. insider trading) the system is constantly on the lookout for things that fall outside of the norm. It does this by establishing profiles of a trader’s behavior, and then comparing those profiles to new behaviors over time, and also comparing the trader’s behaviors to his/her peers.
- Holistic is the key to Identifying Intent. Instead of using siloed analysis and relying on metrics like P&L, SURVEIL-X uses multi-dimensional analytics to uncover true intent. By correlating trades, eComms, voice communications and other related data streams to create a sequence of events, SURVEIL-X helps firms put actions into context to uncover the true intent behind them.
If you have any questions on this topic, feel free to email me at Anurag.Mohapatra@nice.com.