Speech
Introduction
When I spoke to you last year, I reflected on a series of record-breaking highs we had seen in global equities markets, and the corresponding impact that had on prime brokers’ balance sheets. Those rising equities valuations mechanically translated into higher prime financing balances. As funding capacity became more constrained, spreads widened significantly in wholesale equities financing markets. In response to these unique supply and demand dynamics, we saw the emergence of new participants. As hedge funds broadened their prime brokerage relationships to access a finite pool of resources from the banking sector, I highlighted the need for existing prime brokers not to overreach, and for new entrants to proceed with caution in their client adoption and growth plans.
Well, it’s 12 months on, and equities have continued to break further records, albeit with some sharp volatility in early April. Banks have reported prime brokerage balances to be at new peak levels. The S&P finished the year another 16% higher - driven again by mega cap technology stocks. This has led to widespread concerns of a bubble in AI related stock valuations.
I am afraid that I cannot offer any more insights than those that have already been given into whether there is a bubble, or if and when it will burst. However, I can confirm that we are closely focused on the risks and potential consequences for banks if such an event were ever to happen. So, I can offer my perspective on where some of the critical vulnerabilities may lie if a sudden, disruptive sell off in risk assets were to occur.
Technological Advances
Rapid technological advances, such as quantum computing and Generative AI, and their broad applications in our lives, are central to the topics I intend to discuss today. Machine learning and AI is now used by many market participants to process and analyse vast amounts of data to gain a trading edge. But alongside these developments, we have seen major advances in infrastructure and other hardware capabilities. Through the intersection of both software and hardware innovation, transaction execution times have fallen dramatically over recent years. Microwave technology and signal towers have replaced fibre optic cables as the order transmission channels of choice for high frequency traders. Messages and orders pass across these low latency wireless networks - on a line-of-sight between trading centres - at close to the speed of light. Trade execution latency, that used to be measured in milliseconds, is now measured in microseconds, and even in nanoseconds.
A changing market landscape
Whilst new technologies have reshaped markets since the Global Financial Crisis, regulation has had a transformational role too. As banks adapted their business strategies after the crisis, moving away from product focussed proprietary trading models to more client-oriented businesses, driven in large part by regulatory changes, Non-Bank Financial Institutions moved into the space that banks vacated. Non-bank electronic market makers and ultra-low latency traders (or Principal Trading Firms) became established liquidity providers in many vanilla products. Some of these firms are now household names, grabbing headlines from their outperformance and rapid revenue growth. The largest Principal Trading Firms now generate the same trading revenues as the major global investment banks.
On the face of it, the proprietary trading risks that post-crisis global reforms to bank regulation were designed to mitigate have moved from highly regulated banks to the non-bank sector. Instead of depositors’ funds being placed in harm’s way, it is now the private capital of shareholders and founders of these Principal Trading Firms that is exposed to the risks and rewards of proprietary trading. Or so it would seem.
“Frenemies”
Although the major investment banks shuttered their proprietary trading desks many years ago, they still operate as market makers, facilitating client business in many products. But as these products have increasingly moved onto electronic trading venues, Principal Trading Firms now compete directly and fiercely with banks in market segments such as FX and credit. In these largely principal-based markets, where banks have historically dominated through their commitment of capital, a battle for supremacy is raging. In other, ‘agency style’ or commission-based markets such as cash equities and equity ETFs, Principal Trading Firms have largely displaced banks as the core market makers on organised venues. However, Principal Trading Firms do not only compete with banks, they also rely on banks for financial leverage, clearing, treasury and payments facilities, and market access. So, whilst they are competitors, they are also clients. In this regard they can be regarded as “Frenemies”. As some of these Principal Trading Firms have grown in size and importance - in large part through the support of banks - they have in fact become very valuable clients.
Risks have not entirely left the banking sector, they have just been transformed
Now this is where the risks from proprietary trading come back to the banking sector, albeit indirectly. As I have said, Principal Trading Firms need leverage from banks to operate their market making strategies, and they need the good credit standing of banks to access those markets and their associated clearing services. In many cases, they employ the same bank to satisfy both needs at once. The market risks that post crisis regulatory reforms displaced from the system have therefore been transformed into counterparty risks.
For instance, many Principal Trading Firms use banks’ trading infrastructure to execute and settle a portion of their market-making activities in listed derivatives and cash equities. They depend on banks to facilitate and intermediate their OTC FX spot and derivatives transactions. Certain banks who provide services to Principal Trading Firms act as their general clearing members, intermediating their counterparty credit exposures with Central Counterparties (‘CCPs’) for transactions in cleared products, including those executed independently of the bank’s own trading platforms. Banks may, in parallel, act as prime brokers or repo counterparties for Principal Trading Firms, providing financing for their end-of-day positions whilst also supporting their daylight trading activities.
Recently, several Principal Trading Firms have expanded their footprint to encompass activities beyond intraday liquidity provision in specialist products. Several, for example, now run quant-based hedge fund-like strategies with longer holding periods alongside their traditional market making business, or serve as risk re-cycling counterparties to banks for larger, or more complex positions. But their overnight portfolios—financed by prime brokers— have similar risk characteristics to those of hedge funds. As such, the standard risk measurement and risk management tools used by prime brokers for their hedge fund clients’ financing exposures are also applied to these Principal Trading Firm clients.
Growth in intraday exposures
Whilst banks’ end of day exposures to hedge funds and Principal Trading Firms display certain similarities, Principal Trading Firms differ from all other clients due to the sheer volume and velocity of trades linked to their intraday liquidity provision and market making activities. As their servicing banks effectively provide settlement guarantees for these transactions, counterparty risks associated with these intraday trading exposures have risen in lockstep with the growth in Principal Trading Firms’ activities. At the same time, the speed of their trade executions has increased.
Pre-trade versus post-trade controls
When Principal Trading Firms route market orders through the trade execution platforms of their servicing banks, those banks are able to implement automated pre-trade controls and risk limits on their transactions. Such controls protect the banks from taking on client trading exposures beyond their risk appetite. Critically, these protections prevent further orders from being sent into the market immediately upon a limit being breached.
But Principal Trading Firms also use their own direct connectivity to execute trades on exchanges and organised venues in their own right or to transact under bespoke bi-lateral relationships. In these cases, they bypass any pre-trade controls of their servicing banks, whilst the banks remain responsible for settlement of their transactions. Here, their servicing banks rely on the Principal Trading Firms’ own pre-trade controls and on any gateway risk management controls imposed by the trading venues and CCPs. Servicing banks can set up client level pre-trade risk limits at these market platforms or contractually agree aggregate exposure limits with third party execution brokers and bi-lateral counterparties who “give in” trades for them to settle and clear - for example in the FX market. Once the pre-defined risk appetite for a client has been reached, no further trades are accepted by the venue or executing broker. These gateway limits, however, may not be risk sensitive. At some venues they are based on the gross notional value of cumulative orders, without consideration for the directional nature of positions. Even when gateway controls at individual venues provide for net exposure limits, banks often over-allocate their risk appetite across these platforms to give clients enough flexibility to trade freely. So, while this functionality provides some level of comfort, it is far from perfect. In other cases, this functionality does not exist at all, or there are other uncertainties or impediments to banks’ ability to reject trades that are outside their risk appetite.
This whole area is incredibly complex, and the devil is in the detail. But it means that, in many cases, banks are heavily reliant on post-trade monitoring and controls.
Trade execution latency and the impact: response gap
This brings me back to my key theme of rapid technological advancement. When trade execution speeds are measured in microseconds or nanoseconds, placing such reliance on detective, post-trade controls as opposed to preventative, pre-trade controls poses a real challenge for banks. To aggregate positions across venues, banks receive post-trade “drop copy” notifications from Central Counterparties, platforms and brokers. In certain markets there are industry utility providers who facilitate the collection of these reports and calculate client level post-trade aggregate exposures in near real-time. But the process of receiving notifications and aggregating risk exposures, sometimes from multiple different sources, can take seconds, and in busy periods, tens of seconds. Banks set automated escalation alerts for patterns of abnormal client activity based on these aggregated exposure reports. These escalations trigger human intervention. But response times - from trade execution to human action - can be measured in minutes. In some markets where trades are “Given in” to a bank at the end of the day, the time lapse can even be hours. It doesn’t take much imagination to think what damage a clearing client’s misfunctioning, or badly coded, trading algorithm, firing off trades in nanoseconds, can do in that time lapse.
In an emergency, reach for the “Kill Switch”
Upon identification of a problem, you would need to hit the “Kill Switch”. But what is a “Kill Switch”? Using the analogy of a household emergency, it is the stopcock that shuts off the water source when your pipes burst. It is immediate, and it shuts down the source of the problem in one single reflex movement. But what if the flood in your flat is caused by your upstairs neighbour’s pipes bursting? You need to phone your neighbour, or hammer on their door. You need them to act quickly to avert your own potential disaster; you can only hope and rely on them being at home.
Where a Principal Trading Firm accesses markets through trade execution channels hosted by its servicing bank, the bank can shut down all order flow instantly. But how does it work when the Principal Trading Firm executes orders directly into the market, away from the servicing bank guaranteeing its trades? The answer here is more complex.
In the event of such an emergency, servicing banks are initially reliant on the Principal Trading Firm detecting the issue and shutting down its own trading systems at source. If, however, it is the servicing bank who first detects the issue through its post-trade monitoring processes, risk managers would need to contact the client directly to halt order flow and potentially activate any available remote kill switches across multiple venues should those platforms’ controls not already have been independently triggered.
Suddenly many more valuable nanoseconds have been lost, and the bank has guaranteed the settlement of many more trades.
In such an event, the servicing bank would be responsible for the settlement of a significant number of transactions which may need to be unwound at a material loss should the market move against those positions. The bank would then need to recover this amount from the Principal Trading Firm.
Not an abstract risk
Is this scenario genuinely a concern though? After all, Principal Trading Firms’ broker dealer entities are regulated in their own right. They employ leading edge technologies and can attract many of the brightest and the best in their fields. Every message, quotation, or order a Principal Trading Firm sends to the market is under its own direct control and can be automatically monitored by it, in-house. Regardless of execution speed, Principal Trading Firms can incorporate robust pre-trade controls into their models and trade execution platforms, similar to the comprehensive checks and balances implemented by banks. So, I would say that such a scenario is low probability. But it is, nevertheless, potentially high impact.
Over recent years there have been examples of cyber-attacks against financial institutions and financial infrastructure providers, often very sophisticated ones, and in some cases linked to state actors. But most saliently, when Knight Capital, an electronic market maker in the US, released new code within its trading systems into production in 2012, it triggered a continuous stream of erroneous orders onto the New York Stock Exchange. The event reportedly led to losses of almost $500mn and major disruption to the market. This wasn’t a sophisticated cyber-attack; it was a simple version control issue.
So, whilst the instances are extremely rare, the risks of a runaway or rogue algo are not purely theoretical – this has happened before and could happen again.
What should banks do about this?
I’ll now turn to what banks can, and should, do about these risks.
Firstly, anyone moving into this space for the first time should do so with their eyes wide open. As the scale and relevance of Principal Trading Firms has grown, these firms have expanded their banking relationships to secure wider market access and financial leverage. But this is not an aeroplane that you can build as you are flying it. The degree of sophistication needed in risk management systems and approach is extremely high.
The more linear the risk, the more straightforward it is to measure and understand. But whilst simple, near real-time post trade performance and net positioning computations may be relatively straight forward to produce for certain linear products such as cash equities and spot FX, they are much more complex calculations for other asset classes such as commodities that have a calendar risk profile. The challenge becomes even harder for non-linear, option-based products. So, how effective is your intraday counterparty risk monitoring for cleared options portfolios - for example, how frequently do you revalue these positions throughout the day?
There will always be a time lag between trade execution and post trade risk aggregation. But have your own systems and operational capabilities kept up, matching advances in trade execution latency and compute power, or at least not drifted too far behind?
And quite simply, the basics are important. Do you have a comprehensive inventory of every touch point, every asset class/ product type/ and venue, which highlights the effectiveness of pre-trade controls, or where reliance is placed on post trade monitoring? Have you identified all contractual protections and legal considerations? Do you have a playbook should something go wrong? Is accountability and responsibility clearly defined and properly articulated? Have you practiced for every plausible eventuality, and stress tested all possible outcomes and responses?
As the business models of Principal Trading Firms evolve, and the market microstructure continues to change you should also consider whether these intraday risks are growing or becoming more complex. We see developments in a number of markets where Principal Trading Firms have broadened their approach, streaming prices directly to bi-lateral counterparties, moving their focus off trading venues. The more relationships these clients have, the more direct APIs they set up, the greater the response challenge for their General Clearing Member banks and prime brokers becomes should something go wrong – the playbook gets bigger and more complicated. And, as we see the shift within the industry towards cross product netting and portfolio margining, intraday aggregate risk computation becomes more complex and potentially takes more time. Time is critical.
Client due diligence and disclosures
I mentioned in my speech last year that we were focussed on banks’ client onboarding due diligence and disclosure practices and ongoing credit risk assessment processes. The Basel Committee and the Financial Stability Board have both also addressed this topic. I am pleased to say that we have seen good progress amongst banks towards improving their client due diligence risk disclosure frameworks, linking the quality of client disclosures to risk appetite and the degree of financial resources provided. However, many of these developments have been in relation to hedge funds and family offices. We have seen that, in several cases, Principal Trading Firm clients do not fit these internal client categorisation definitions and therefore currently sit outside of those frameworks. As I have highlighted, many of these Principal Trading Firm clients have end-of-day financing positions at their prime brokers that are somewhat similar to those of hedge fund clients. Additionally, as noted, they demonstrate a distinct intraday trading risk profile. Unlike hedge funds, these clients do not have third party investors and therefore they do not produce monthly NAVs or issue investor letters. The amount of information made available publicly by these clients is scant. Whilst these firms have regulated broker dealers in their groups, they may also make use of unregulated group companies to hold risk positions. It is imperative therefore, that banks require at least similar levels of risk disclosures from Principal Trading Firms as they do from hedge funds to manage their exposures to these clients.
The adequacy of disclosures should be assessed in relation to the particular risks that banks face. With respect to Principal Trading Firms, these risks are not solely about the financial health and trading profile of those clients. Banks who are also exposed to intraday risks from Principal Trading Firm clients should satisfy themselves that their clients’ trading systems controls, including relevant technology change management processes, are operationally sound and robust.
Boards should be aware
Boards and their risk management committees need to be inquisitive and aware of risks from daylight exposures to Principal Trading Firm clients and seek assurances that control frameworks are adequate and properly resourced. There is a direct link between risk mitigation and the level of investment in risk management technology, and this should remain a keen focus for executive management and boards. The good news is that current advancements in ultra-low latency trade execution speeds may be reaching their limits. For something that already travels between two fixed points at close to the speed of light, the laws of physics suggest that it cannot get any faster. So, banks are now able to narrow any latency gaps in their risk management controls rather than watch them widen.
The PRA will this year carry out more work with banks on the adequacy of their counterparty risk management frameworks, specifically directed at their control of intraday risks.
I will finish by quoting Benjamin Franklin,
“You may delay, but time will not”.
Thank you.
