In the last few years, following the financial crisis, stress testing has been put in front of the

scene as
an effective tool to ensure soundness of the banking system. Although the 2009 EU-wide stress test
was not well received by the investors community (it lacked some disclosures and a few banks which
passed this exercise were then bailed out by EU and IMF a few weeks later), stress testing is now
seen as a comprehensive and transparent solution to clean up balance sheet and prevent financial
crisis.
The US Federal Reserve now runs every year the Comprehensive Capital Analysis Review and in
Europe, ECB starts a similar process with the Comprehensive Assessment, which includes the Asset
Quality Review and a stress testing exercise.
Obviously similar initiative is done in the UK, and in October 2013, the Prudential Regulation
Authority (PRA) set in a discussion paper a "Framework for stress testing the UK banking system
1
".
This paper drew the guidelines for scenario modelling, asset quality review and stress testing. As part
of this exercise, banks are supposed to send to the PRA their actual data for different type of risk
through what we call the Firm Data Submission Framework (FDSF). Standard templates are designed
by PRA in order to provide a comprehensive framework for banks to submit data.

Implementation Update
The first FDSF exercise occurred in the first quarter of 2014, based on the banks actual data as of
December 31st 2013. The 8 biggest UK banks (SIFIs) were concerned by this first exercise. This was
quite a challenge as the timeline was quite short (only a few weeks to deliver the data) and the
templates were delivered quite late by the PRA.
PRA has provided templates for Retail Credit (with and without mortgages), Wholesale Credit,
Structured Finance and for Projections as well. New templates have been provided lately for
projections of Balance Sheet, Profit & Loss and for Pension risk. They will be used for the submission
of June 30th 2014.
In the coming years, when FDSF runs at full capacity, more than 50 banks will be concerned by the
data submission, and most of them should start preparing their data infrastructure.

Challenges
The FDSF process required by PRA is quite ambitious. The requirements in term of frequency and
timing are strong while in the meantime data granularity and data quality must remain perfect. As
the Association for Financial Markets in Europe (AFME) mentioned, these requirements are even
more ambitious compared to US Fed's CCAR exercise. FDSF leans more towards Risk Management
rather Stress Testing
2
.
FDSF requirements are always evolving. Templates has been sent to the banks throughout the year
of 2013 and new updates are still delivered on a regular basis. This is quite a challenge for banks to
adapt their data infrastructure in order to comply with the latest templates.
PRA Stress Testing is a yearly exercise, but the data submission is done on a quarterly basis for most
reports.
The templates to provide quarterly are the following:
Retail Mortgage Credit Risk
Retail Excluding Mortgage Credit Risk
Corporate, Sovereign and Institutional Credit Risk
Asset and Liability Management / Liquidity Risk
Structured Finance
Counterparty Credit Risk
Operational Risk
Pension Risk
The Market Risk template must be provided monthly. Projection template must also be provided.
PRA may also require ad-hoc submissions of these reports when it announces a new stress test or
when a bank has significantly changed any of the information typically submitted annually
(restructuring, merger & acquisition, change of business model...).
The quarterly submissions must be "as of" the end of each calendar quarter and the data must be
submitted no later than 30 working days after the reporting date. For end of year submissions, this
delay is extended to 50 working days. This is quite challenging given the granularity and
comprehensiveness of the data.
FDSF Actual reports to provide are very granular, even much more granular than EBA COREP reports.
The information is to be reported per portfolio (commercial entity, business unit, country, asset
class...). Counterparty level information is required for top 20 counterparties per asset class. Some
data must also be populated manually given the required frequency and the lacks in some legacy
systems, therefore making even harder the data quality process.
All actual FDSF reports must also be consistent and reconciled including with information reported in
other more aggregated reporting stack (e.g. FSA004, FSA045, COREP, FINREP ). Risk and Finance
departments must integrate their data and models (e.g. ensure consistency between RWA forecast
and ALM Balance Sheet growth). Also, projections must also reconcile with actual baseline data from
the first quarter. This will again require an efficient data infrastructure.
Banks also need to be prepared to benchmark their models. Indeed, PRA could perform its own
projection and stress test using the actual information reported and then it could challenge actively
banks results and models. Also, PRA may require banks to back test their results, wondering for
instance "why Year 2 actuals not in line with Year 2 projections performed Year 1?".
Data quality is a critical challenge for FDSF, as PRA may take conservative view on stress test if it is
not satisfied with the quality of the data submitted by banks. Indeed, one of the objective of the
FDSF initiative is to enhance data quality for stress testing and therefore improve the quality of data
used by the senior management of the banks.
The supervisor assesses through automatic tools the quality of all structured data submission (either
in Excel or XML format). These tools use business and technical rules. For instance, they check that
mandatory fields are populated, that data types are correct, that only authorized values are used,
that reports reconcile altogether, and that amounts are consistent (i.e. EAD, PD, LGD and Expected
Loss must be consistent). If the submitted data fails these checks, an error is reported and the
submission rejected. The PRA then checks the plausibility of the submissions, from a business point
of view, for instance assessing how the data evolves over the years and if it is consistent with the
business of the bank.
The PRA also requires banks to fully document their data submission process and the data
provenance flow. For instance, a Data Flow Diagram for each type of risk should be maintained by
each bank. It should mention the provenance of material data for each submission. PRA insists that
this documentation should be readable by non-IT specialist, therefore an effort is done in the way
the data process is documented. Any transformation of data, aggregation, merging, mapping, ... is
therefore documented and clearly marked.
All the challenges above occur in a time when PRA's FDSF is not the only supervisor requirement for
UK banks. Indeed, for international banks, other local supervisor may have similar requirements at
the same time. Also, the EBA-ECB Comprehensive Assessment may have an impact for UK banks and
require time and effort. Therefore, there is a shared feeling in banks that several supervisors ask the
same kind of work twice or more and that some of it could be rationalised if more coordination
happens between supervisors. For instance, supervisors should ensure that the reporting date used
for different data submissions are the same.
Banks may have taken tactical decisions for the first FDSF submission in 2014 but FDSF will be here
for a while and will still require more intensive systems (to ensure quality and granularity of the data
in a frequent manner). Therefore, strategic implementation must be done and banks must find a way
to reconcile tactical solutions with the long term strategy of the bank. That is a main reason why the
involvement of senior management is paramount (beside the fact of course that each FDSF
submission may have a significant impact on bank reputation).

Beyond FDSF
In January 2013, Basel Committee on Banking Supervision (BCBS) issued a set of principles to
strengthen banks risk data aggregation and risk reporting practices (BCBS 239
3
).
The FDSF exercise should be thought as a step towards the implementation of those principles.
Indeed, BCBS 239 states that Global Systemically Important Banks (G-SIBs) must implement these
principles by 2016 and as of the beginning of 2014, FDSF involves UK G-SIBs.
BCBS 239 highlights how bank's data infrastructures should improve in the coming years, to ensure
better data quality and therefore better risk management and senior decision. All of the FDSF
requirements and challenges seen above are in a way described in BCBS 239. Therefore,
implementing the Risk Data Aggregation principles will ease FDSF exercises.

1
http://www.bankofengland.co.uk/financialstability/fsc/Documents/discussionpaper1013.pdf
2
http://www.afme.eu/WorkArea//DownloadAsset.aspx?id=10200
3
http://www.bis.org/publ/bcbs239.pdf