Speech
Good afternoon and thank you for inviting me today.
You’d be forgiven, when you heard I’d be talking about the PRA’s approach to data collection, for thinking of spreadsheets. Enormous spreadsheets, probably whizzing in one direction – from firms to their regulator – at great cost and complexity.
There is truth in that obviously, but it misses something important. The point I’m going to make today is that data is a form of dialogue: it’s the foundation of a structured, ongoing conversation between the PRA and the firms we regulate about risk.
The financial system is complex, constantly changing, and too large to be understood through individual conversations alone. Data underpins how the regulatory dialogue happens at scale. It’s the way firms explain, in a common language, how they are managing risk. And it’s critical to how the PRA listens, understands and responds to the risks in the system. Given its importance, how we design, collect, and use data really matters. That is the context for the work of our Future Banking Data (FBD) programme, launched last year.
As those here well know, the PRA’s primary objective relating to banks is to promote their safety and soundness. We also have two secondary objectives, namely facilitating effective competition in the markets for the firms we regulate and facilitating the competitiveness of the UK economy and its growth in the medium to long term subject to alignment with international standards.
Data is central to delivering each of these objectives, and so we collect a wide range of data via management information, one-off requests and, critically, regulatory returns from firms.
But while talk may be cheap, as they say, data is not, and so the FBD programme aims to deliver a transformative reduction in banks’ reporting costs, as well as substantive improvements to the relevance, quality and timeliness of our data collections.
Different kinds of conversation
In any relationship, we need to have different sorts of conversations at different times. The conversations we have with you as your regulator – about supervision, market wide risk assessment and policy development – are no different. Each of those conversations is grounded in data and draws widely from the same data vocabulary.
Regulatory returns
Our Rulebook returns can be thought of as a standardised check-in or scene-setter. They’re the established opening to our conversation – the ‘how are you?’ and the ‘what’s the weather like where you are?’– The data in these returns form the baseline that allows us to understand firms’ balance sheets, exposures and resilience over time. And depending on the answers to those initial questions there might not be any follow up needed or we could be settling in for a long chat.
That baseline informs the PRA’s assessment of risks to safety and soundness across areas such as capital adequacy, liquidity, governance and operational resilience and likewise supports the Bank’s assessment of systemic risks in these areas. Core regulatory returns, including COREP and FINREP, feed into the PRA’s risk assessment framework and support firm level supervisory engagement. This includes key supervisory milestones like Periodic Summary Meetings, where supervisory strategies and priorities are set.
Because regulatory returns are collected on a consistent basis across all relevant firms, they enable benchmarking and cross firm analysis. This allows supervisors to identify outliers, detect emerging patterns and assess whether risks are firm specific or more widespread across the system. Such comparability is an important complement to firm specific observations.
Thematic and ad hoc collections
Thematic and ad hoc collections add flexibility to the conversation, allowing us to ask more specific questions of relevant subsets of firms when we need to explore a particular risk in more depth.
These collections allow supervisors to remain responsive to new issues, rather than relying solely on standing data that may lag behind events. They are an important part of the PRA’s judgement based and risk focused approach to supervision. They enable supervisors to respond flexibly to changes in firms’ business models, market conditions, or the external environment, and to deepen understanding of particular exposures, behaviours or vulnerabilities. In many cases, they draw on firms’ existing management information, recognising the value of data that firms already use to run their businesses and manage risk.
In business-as-usual conditions, such collections are used to supplement regulatory returns and peer analysis, helping supervisors refine their assessment of risk and prioritise supervisory attention.
In periods of stress or uncertainty, the balance can shift further towards ad hoc requests, where timeliness may take precedence over comparability in order to support rapid assessment and decision making.
In both states of the world, we recognise that such collections can carry additional cost for firms, as they cannot always be planned for or delivered through established systems, and are therefore used selectively and proportionately.
Stress tests and scenario data
Stress tests and scenario data add a different dimension. They are the ‘what if?’ questions in the conversation – explicitly forward looking and designed to explore how firms and the system might respond under adverse but plausible conditions. As with our regulatory and ad-hoc collections, scenario data can cover both financial and operational risk and resilience.
Through the Stress Test Data Framework, participating firms provide more granular and scenario-based data on capital adequacy, credit exposures and other risks that impact financial resilience. Firms’ projections are analysed against peer results and internal benchmarks, supporting judgements about asset quality, capital adequacy and the credibility of firms’ assumptions under stress. These data support forward looking assessments of resilience and inform macroprudential policy judgements, including those related to capital buffer setting.
Each of these types of questions and answers matter but it is when the different parts of the data conversation come together that their value is most apparent. My own area of the PRA, Prudential Policy, is a good example of this. Policy development is particularly data intensive, especially in the calibration, impact analysis and cost benefit assessment of the options. For similar reasons, research, which underpins our understanding of market failures, the impact of existing policy measures and can help us develop new solutions, can generate very particular data demands.
Both policy and research often require high-quality time series data, either to assess outcomes before and after implementation, or to construct econometric counterfactuals. Firm-level data allows us to evaluate whether the policies we have already implemented are working as we intended, and if not, what needs to be adjusted.
Of course, dialogue is most effective if we listen carefully to what the other party is telling us. Although data often speaks for itself, a judgement-led supervisory model means that there is value in taking time to challenge and test interpretations. And it should go without saying, that poor quality data distorts the conversation.
Recognising the challenges
Any conversation can become unproductive if it’s too noisy, too frequent, or disordered. That’s why getting the right data, in the right way, at the right time matters: the quality and focus of the dialogue matters more than the volume.
As I’ve outlined, the data collections we receive today are of enormous value, driving constructive interactions with supervisors, policy teams, and colleagues across many other areas of the Bank. Nonetheless, we have heard – including from many here today – that there is room for improvement.
One of the challenges we have heard from firms is that, as they prepare data to send to us, it would help to have a better understanding of how we plan to use it. Which is why, in our recently published FBD discussion paper we included detailed case studies that explore how we use data for firm supervision in business-as-usual circumstances, for crisis management, for cross-sectional analysis, and for making policy.
We’ve also heard feedback from firms that our reporting requirements are hard to understand in places. While we make substantial efforts to ensure instructions are easy to follow, we recognise that where firms face challenges in identifying what to submit, data quality can be lower. So, we’ll be working to be clearer in what we ask.
Another challenge is the volume of our requests. At present we collect over 400 banking data templates across the Bank and PRA. While no one firm is required to complete all of these templates and there is little outright duplication, collections sometimes ask for similar underlying data to be cut or aggregated in different ways. There are also legacy templates that may have lost some value, either through being superseded or changes in circumstance.
We have already made a start on removing or reducing collections in areas that have declined in relevance. In December of last year, we completed our first phase of deletions doing away with 37 templates, mostly from FINREP, that did not provide sufficient supervisory or policy value to justify us continuing to collect them.
We have also made reporting for smaller firms more proportionate under the Strong and Simple framework. And my insurance colleagues would want me to point out that in our recent Solvency UK reforms we cut regulatory reporting for insurers by a third. Put bluntly, we don’t want to waste time and effort – yours or ours – on things we already know or just aren’t of relevance anymore.
But, just as importantly, having the right data also means closing data gaps where we see them. While our data estate has been continuously amended over time, the risk environment has changed even more. Stress events and crises – which seem to be happening more frequently in recent years – can often reveal gaps where data are required but are not currently reported. One such example is Archegos; which highlighted data gaps that constrain financial stability monitoring. More recent stresses and developments in private markets have underscored the need to add new topics to our data conversations. In time, we expect to address these gaps with targeted new collections.
Nonetheless, consistent with the overall aim of reducing the burden on firms, the goal of the FBD programme is better focused data requests which can be both adapted to changing needs and answered more efficiently.
The changes we need to make
To guide broader work on data over the coming years, our discussion paper suggests four principles, which build on current PRA practice. We want our data dialogue to:
- Be anchored in the PRA’s objectives;
- avoid repetition by collecting data ‘once and well’;
- be easy for firms to engage with; and
- evolve to remain fit for purpose over time.
The substance in these principles is not new: for example, all rules made by the PRA, including for reporting, are underpinned by its objectives and supported by cost benefit analysis. Nonetheless, by setting out these principles for discussion, we hope to improve industry understanding of our approach and be transparent about how we intend to deliver cost reductions and other improvements.
Every data collection has different characteristics, and in designing them we need to weigh-up a range of trade-offs. To pick just one example, we know many of our firms also report to regulators in other jurisdictions. While we recognise that aligning data collections with overseas authorities might benefit some, simpler returns could benefit a more diverse range of firms. And just as we need different data for different kinds of conversation, the PRA will seek to adjust its data vocabulary to best suit the UK financial system.
Finally, we need to make sure our conversations are grounded in the right level of detail. There is a trade-off to be made as we balance the collection of granular data versus aggregates. Granular data collections, at the exposure level, for instance, offer greater potential for reuse because they can be cut and recut for different purposes. While they may carry additional up-front costs, reuse may ultimately permit a lower overall reporting burden.
Work under way
While feedback on the discussion paper will steer our future work, we have a number of initiatives in train already.
First, in keeping with our aim for a substantive, meaningful data dialogue, we will continue to dial back on unneeded noise. Given the unanimous support we received to our September consultation regarding the first phase of reporting deletions, we are pressing ahead with identifying further templates that are underused relative to the burden they impose.
Second, we want to make our lines of communication as easy to use as we can. Last month we launched testing of a new communication portal with select firms to better gather data needed to support regulatory transactions. This firm-facing portal will facilitate interactions with the PRA, focusing on improving firms’ experience with data collections.
Initially, the portal will make it easier for firms to share data for certain authorisations purposes. Testing of the first stage of the portal is already under way, and we have the intention to roll this out later in the year for banks submitting a range of permission applications.
Third, recalling my comments on granular data, mortgages are a perennial conversation for many areas of the Bank and PRA. This is one area we are persuaded it may make sense to collect data at the exposure level. The chart in the annex illustrates the case: Mortgage data are one of the most intensively used themes across the Bank and PRA, feeding into several of our responsibilities.
These are highly valuable data; the multitude of regular uses demonstrates the importance clearly – and this chart ignores a wealth of one-off analyses and research. But the plethora of differing collections also speaks to the cost. The chart shows around 54 mortgage data categories spread across 11 separate mortgage-related data collections, used across 9 different areas of the Bank and PRA. Together, these data feed into more than 70 end-user tools, reflecting the breadth of mandates and objectives they support.
Therefore, we are currently undertaking a feasibility study that will assess whether and how we could integrate the 50 or so collections across the Bank and PRA by expanding on some of our existing loan-level collections, drawing lessons from international peers.
We will need your help to do this and will soon be launching an industry group of mortgage data experts, which I'd encourage you or your colleagues to join.
While these examples will generate tangible progress, I would like to keep expectations in check. Given the scope of the PRA’s banking data and interlinkages with the Bank and FCA’s responsibilities, it is our view that a ‘big bang’ approach would present excessive up-front costs and delivery risks. Hence, we are proposing an incremental approach, which would allow us to be agile, flexible, and course correct as we progress. Equally importantly, we think that approach should enable firms to gain some benefits sooner.
Conclusion
Ultimately, the purpose of our data collections is to support the PRA in identifying emerging risks early, acting proportionately, and reducing the likelihood and impact of firm failure.
To ensure that these collections are, and remain, fit for the future and are not unnecessarily burdensome to produce we are revisiting how we use the data we gather in practice. This includes identifying the challenges with the current framework and working out the principles and trade-offs that will guide our approach to reform.
Through this work – and in keeping with my theme of dialogue – runs an emphasis on doing this in partnership with firms. We encourage you all to share your views on the issues raised in the discussion paper, including the principles for future reporting, the trade‑offs involved, and where there may be the biggest opportunities to reduce firm burden and therefore costs.
And now that my dialogue metaphor has itself become a bit meta, as I direct your attention once again to a discussion paper about a discussion, I’ll wrap up by simply saying that we look forward to continuing this work collaboratively with firms and stakeholders over the coming months. Keep talking to us!
I would like to thank David Bailey, Mike Burke, Jas Ellis, Imeth Kapuruge, Hana Addow, Elena Lombardi, Aaron Shiret, and Dane Whittleston for assisting me in preparing these remarks.
Accompanying slide