An inconsistent approach to CET1 remediation in the Asset Quality Review

The ECB’s Asset Quality Review is a huge project (reviewing EUR 3.7 trillion of risk weighted assets at 128 banks) and hence it was big news when the ECB published its manual with instructions for the local inspectors for the execution of the review. The ECB aims to review the assets of EU banks before it will assume supervision in November 2014.

The AQR is phase 2 of the process and will be finalized in August 2014. The phase ends with the calculation of the AQR-adjusted Common Equity Tier 1 (CET1) ratio. The final phase will be a stress test. Banks are required to comply with a CET1 minimum level of 8% and 5.5% during stress.

This phase 2 consists roughly of a submission by each bank of transaction information (the ‘loan tapes’), the data integrity validation (DIV) of these data and an in depth review of the provisioning and fair value calculations. The corrections after this in depth review will result in the AQR-adjusted CET1 ratio.

A hastily written manual for an AQR process under time pressure

The process as a whole is under severe time pressure and this applies to the manual as well.

The manual lacks annexes containing methodological explanations, reference lists and a glossary (not a luxury item given the many abbreviations used). The chapters contain sentences that seem to have been quickly written down including typos. An example is (my italics, p.33):

Using Basel EADs will be definition be conservative, therefore in these circumstances, bank’s will be allowed time following the PP&A to provide CVA exposure profiles for trades not initially included in the exposure profile.

Formulas are not always clearly stated, see e.g. the expression of the relationship between Loan-to-Value and Loss Given Loss that the bank is supposed to fit:

Formulas are not always written down in a neat way

Formulas are not always written down in a neat way

More haste is evidenced in the DIV approach: on the one hand, the manual explains that (p.35, my italics):

Furthermore, any analysis to be performed concerning potential capital shortfalls and stress testing is predicated on a thorough understanding of the data issues. As a result, a thorough DIV approach is required.

That is twice the word ‘thorough’ in two subsequent sentences.

On the other hand the manual says (p.36):

The DIV process should be a “straight line” process that is executed within the time allowed. It should not result in continual reiteration of loan tape data. Instead, appropriately conservative remediation strategies should be applied to avoid the need for delays in the process while data is corrected

The DIV should be carried out within a month, from 14 March 2014 till 11 April 2014.

Non performing exposures, Defaults and Impairments

The AQR adheres to the EBA simplified definition of a non performing exposure (NPE), sc. a loan that is either 90 days past-due, and/or impaired and/or in default. (p. 46)

The defaults seem to be a subset of the impaired portfolio, since default is one of the impairment triggers (table 38 starting on p.114). However, we cannot be 100% sure since we cannot rule out that different cure definitions may lead to an outflow out of the impairment status while the loan is still in default. (Thanks to colleague dr. Sven de Man who pointed this out to me.)

Adjustments of provisions and fair value are deducted from available capital

A credit file review will be carried out. Individual transactions will be sampled and assessed for accuracy of the provisioning if a specific provision is required.

The findings of the credit file review will be projected to the wider portfolio. This may lead to an additional provision that will feed into the AQR adjustment of the numerator (a deduction from Common Equity capital) of the CET1 ratio.

The manual explains that provisions are fully deducted from available capital (p. 260) and the ‘AQR-adjusted CET1%’ will be adjusted for deviations in estimates of provisions and level 3 valuations (Credit Value Adjustment) that result from the AQR. (p. 260)

The denominator of the CET1 ratio, i.e. the risk-weighted assets, is not adjusted. The manual says about these RWA:

For the purposes of the AQR-adjusted CET1% RWA will not be adjusted given materiality and in the interests of feasibility of the exercise, except for the impact of change to level of protection from risk transfer transactions/securitisations etc. Of course, once adjustments to accounts are made following the completion of the CA, the associated adjustments to RWA would be expected to be made by the bank.

Adjustments to specific and collective provisions

The sampling procedure for the correction of the specific provisioning of impaired exposures is stratified across ‘common risk strata’. These strata are defined with the help of risk parameters that do not rely on PDs and LGDs, such as ‘default more than 12 months’, ‘high risk cured’, ‘high risk normal.’ The ‘high risk normal’ stratum contains loans that are identified with the help of criteria such as Debt / EBITDA > 6, the loan is on the watchlist or Loan-to-Income > 5 (for retail mortgages). (table 28, p. 82).

Note that the stratification for the sampling procedure is not based on PDs and LGDs.

Within the common risk strata, the loans are divided across size buckets based on exposure and not on e.g. loss at default. Again, the LGD is not used within the analysis.

The sampling from the size buckets makes the procedure effectively a Monetary Unit Sample in which the probability for a loan to be selected is proportional to its size (ppswor sample: probability proportional to size without replacement). The most likely misstatement percentage is correctly calculated as the unweighted average misstatement percentage of the sampled files within the strata of one common risk block. (figure 25, p. 171)

For the collective provisioning of smaller, homogeneous, impaired exposures the bank will use a model (for retail mortgages) with parameters Exposure at Default (EAD), Probability of Impairment (PI), Cure Rate (CR) and Loss Given Loss (LGL). (p. 177)

Since the model should accommodate a cure rate vector of past due states that can deal with state transitions from/to multiple states and four years (p.191), a Markov chain analysis is required here. The analysis will possibly lead to a corrected Incurred But Not Reported (IBNR) figure / general provision.

The emergence period (sometimes referred to LIP, Loss Identification Period) for performing exposures is conservatively set at 12 months (p.177):

The rebuttable presumption will be that a 12-month emergence period will be used for performing exposures.

However if a bank can provide evidence that a shorter emergence period is more appropriate then this shorter period should be applied. (I guess this is what is meant by ‘rebuttable’.) Remarkable since this provides room for maneuvering.

Traditional risk parameters are neglected; however, the RWA part of the CET1 ratio is not adjusted

As we have seen above in the sampling procedure, the traditional risk parameters at the loan level are completely neglected. The comprehensive assessment uses analysis methods that rely on basic risk indicators such as debt/EBITDA, Loan-to-Income, days past due, normal but cured after default, et cetera.

The underlying reason seems to be a lack of confidence in the bank’s PD and LGD calculations. Also, the fact that PD and LGD are through-the-cycle parameters may play a role.

This leads to an inconsistency in the entire exercise. The problem is: the provisioning adjustments and fair value (a.o. CVA) adjustments are only fed into the numerator of the CET1 ratio (deductions from available capital). But the denominator, the risk-weighted assets, is left intact.

That is strange for two reasons: first, if there is an underlying lack of confidence in the bank’s PDs and LGDs this carries over to the RWA. And second, the entire phase 2 will probably lead to new insights on PD and LGD. These insights are not captured in PD and LGD adjustments.


Over Folpmers
Financial Risk Management consultant, manager van een FRM consulting department, bijzonder hoogleraar FRM

Geef een reactie

Vul je gegevens in of klik op een icoon om in te loggen. logo

Je reageert onder je account. Log uit / Bijwerken )


Je reageert onder je Twitter account. Log uit / Bijwerken )

Facebook foto

Je reageert onder je Facebook account. Log uit / Bijwerken )

Google+ photo

Je reageert onder je Google+ account. Log uit / Bijwerken )

Verbinden met %s