IFPRU 4.4 Internal ratings based approach: overall requirements for estimation

In order to be able to determine that the requirements in article 144(1) of the UK CRR1 have been met, the FCA would typically have the high-level expectations set out in this section.

The information that a firm produces or uses for the IRB approach should be reliable and take proper account of the different users of the information produced (customers, shareholders, regulators and other market participants).

A firm should establish quantified and documented targets and standards, against which it should test the accuracy of data used in its rating systems. Such tests should cover:

1. (1)

   a report and accounts reconciliation, including whether every exposure has a PD, LGD and, if applicable, conversion factor for reporting purposes;

2. (2)

   whether the firm's risk control environment has key risk indicators for the purpose of monitoring and ensuring data accuracy;

3. (3)

   whether the firm has an adequate business and information technology infrastructure with fully documented processes;

4. (4)

   whether the firm has clear and documented standards on ownership of data (including inputs and manipulation) and timeliness of current data (daily, monthly, real time); and

5. (5)

   whether the firm has a comprehensive quantitative audit programme.

In respect of data inputs, the testing for accuracy of data (including the reconciliation referred to above) should be sufficiently detailed so that, together with other available evidence, it gives reasonable assurance that data input into the rating system is accurate, complete and appropriate. The FCA considers that input data fails to meet the required standard if it gives rise to a serious risk of material misstatement in the own funds requirement either immediately or subsequently.

In respect of data outputs, a firm (as part of the reconciliation referred to above) should be able to identify and explain material differences between the outputs produced under accounting standards and those produced under the requirements of the IRB approach, including in relation to areas that address similar concepts in different ways (eg, expected loss and accounting provisions).

A firm should have clear and documented standards and policies about the use of data in practice (including information technology standards) which should, in particular, cover the firm's approach to the following:

1. (1)

   data access and security;

2. (2)

   data integrity, including the accuracy, completeness, appropriateness and testing of data; and

3. (3)

   data availability.

[Note: article 144(1)(a) of the UK CRR1)

For the FCA to be satisfied that a firm documents its ratings systems appropriately, in accordance with article 144(1)(e) of the UK CRR1, it would expect a firm to be able to demonstrate that it has an appropriate policy for any ratings system in relation to:

1. (1)

   any deficiencies caused by its not being sensitive to movements in fundamental risk drivers or for any other reason;

2. (2)

   the periodic review and action in the light of such review;

3. (3)

   providing appropriate internal guidance to staff to ensure consistency in the use of the rating system, including the assignment of exposures or facilities to pools or grades;

4. (4)

   dealing with potential weaknesses of the rating system;

5. (5)

   identifying appropriate and inappropriate uses of the rating system and acting on that identification;

6. (6)

   novel or narrow rating approaches; and

7. (7)

   ensuring the appropriate level of stability over time of the rating system.

[Note: article 144(1)(a) and (e) of the UK CRR1)

To be satisfied that the requirements in article 179(1) of the UK CRR1 are met, the FCA expects a firm to collect data on what it considers to be the main drivers of the risk parameters of probability of default (PD), loss given default (LGD), conversion factors (CFs) and expected loss (EL) for each group of obligors or facilities, to document the identification of the main drivers of risk parameters, and be able to demonstrate that the process of identification is reasonable and appropriate.

In its processes for identifying the main drivers of risk parameters, the FCA expects that a firm should set out its reasons for concluding that the data sources chosen provide in themselves sufficient discriminative power and accuracy and why additional potential data sources do not provide relevant and reliable information that would be expected materially to improve the discriminative power and accuracy of its estimates of the risk parameter in question. This process need not necessarily require an intensive analysis of all factors.

[Note: article 179(1)(a), (d) and (e) of the UK CRR1)

To demonstrate that rating systems provide for meaningful assessment, the FCA expects that a firm's documentation relating to data should include clear identification of responsibility for data quality. A firm should set standards for data quality, aim to improve them over time and measure its performance against those standards. Furthermore, a firm should ensure that its data is of high enough quality to support its risk management processes and the calculation of its own funds requirements (see article 175(1) of the UK CRR1).

Further detail of standards that the FCA would expect a firm to meet when it assesses compliance with article 174 of UK CRR1 are set out in the sections on probability of default (PD), loss given default (LGD) and exposure at default (EAD).

In assessing whether the external data used by a firm to build models is representative of its actual obligors or exposures, the FCA expects a firm to consider whether this data is appropriate to its own experience and whether adjustments are necessary (see article 174 of the UK CRR1).

To estimate PDs that are long run averages of one-year default rates for obligor grades or pools, the FCA expects a firm to estimate expected default rates for the grade/pool over a representative mix of good and bad economic periods, rather than simply taking the historic average of default rates actually incurred by the firm over a period of years. The FCA expects that a long run estimate would be changed when there is reason to believe that the existing long run estimate is no longer accurate, but that it would not be automatically updated to incorporate the experience of additional years as these may not be representative of the long run average (see article 180 of the UK CRR1).

To demonstrate compliance with article 144(1) of the UK CRR1, the FCA expects a firm to take into account the following factors in understanding differences between their historic default rates and their PD estimates, and in adjusting the calibration of their estimates as appropriate:

1. (1)

   the rating philosophy of the system and the economic conditions in the period over which the defaults have been observed;

2. (2)

   the number of defaults, as a low number is less likely to be representative of a long run average. Moreover, where the number of internal defaults is low, there is likely to be a greater need to base PDs on external default data as opposed to purely internal data;

3. (3)

   the potential for under-recording of actual defaults; and

4. (4)

   the level of conservatism applied.

The FCA expects a firm that is unable to produce a long run estimate, as described above, to consider what action it would be appropriate for it to take to comply with article 180(1)(a) of the UK CRR1. In some circumstances, it may be appropriate for a firm to need to amend its rating system so that the PD used as an input into the IRB own funds requirement is an appropriately conservative estimate of the actual default rate expected over the next year. However, such an approach is not likely to be appropriate where default rates are dependent on the performance of volatile collateral. (See articles 179(1)(f) and 180(1)(a) of the UK CRR1).

In accordance with articles 181(1)(b) and 182(1)(b) of the UK CRR1, where the estimates appropriate for an economic downturn are more conservative than the long run average, the FCA expects the estimate for each of these parameters to represent the LGD or CF expected, weighted by the number of defaults, over the downturn period. Where this is not the case, the FCA expects the estimate to be used to be the expected LGD or CF, weighted by the number of defaults, over a representative mix of good and bad economic periods (see articles 179, 181 and 182 of the UK CRR1).

To demonstrate that a rating system provides for a meaningful differentiation of risk and accurate and consistent quantitative estimates of risk, the FCA expects a firm would have regard to the sensitivity of the rating to movements in fundamental risk drivers, in assigning exposures to grades or pools within a rating system (see article 171 of the UK CRR1).