24 May 2014

Notes on ECB’s Comprehensive Bank Assessment (Part 4)

The second* key pillar in the ECB’s Comprehensive Bank Assessment is a stress test, following the Asset Quality Review (AQR), which is discussed in Part 3 of this series. On 29 April, the European Banking Authority (EBA) released the stress test methodology and scenarios.

EBA faced a difficult choice in calibrating this EU-wide stress test: if they implemented a strict and credible test they risked imposing politically costly capital demands on EU banks; on the other hand, if they implemented too weak a test they would further damage the already blemished reputation of EU banking sector stress tests and (as a commentator has pointedly put it) “look like pointless clowns in the bargain”.

Now the choice has been made and the 2014 stress test scenario is available for commenting. However, as I have pointed out earlier when analyzing the 2011 EU-wide bank stress test results, the scenario may not even be the main issue. How it is being translated into the conclusions at the end, is just as if not more crucial. Therefore – and as other observers mostly focus on the scenario – I’ll explore the Methodological Note first.


Review of 2014 EU-wide bank stress test methodology

Working they are hard in the EBA, no doubt about that. Methodological Note itself is a 70 pages document – and surely not an easy writing but including two pages of abbreviations, a number of brain-teasing formulas and calculation rules, references to 24 core and 12 additional Excel templates for the banks to fill etc.

Before getting lost into the details, there is a stupid (or not so stupid) question to be asked and answered: Why all this complexity? Isn’t banking just about taking deposits and granting loans? It should be pretty trivial…

But it’s not:
  • Banking business itself has become much more complex than that. As an illustration, 50% or more of the large euro area banks’ balance sheets consists of other often hardly comprehensible instruments than old-fashioned loans and deposits; at least one fifth of their assets (if these are actually assets – but that’s yet another question) are held for trading. Valuation, let alone forecasting this toxic mix consistently very much looks like mission impossible (especially given that banks are not operating in isolation, but the whole financial system is extremely interconnected). 
  • Regulators are “eating their own cooking”: banking regulations have enabled running a bank with almost no capital by justifying it, surprise-surprise, with lots of complexity.** If the EBA now presented a fairly simple framework and said that the simple non-risk based leverage ratio after the stress test should be let’s say 5.0% (max 20x leverage wouldn’t sound such a unreasonably conservative expectation – other way round),  virtually every if not every major bank in the exercise would fail.  Therefore they have chosen another route: testing banks’ solvency against the CRD IV/CRR rules, i.e. on the basis of complex risk-weightings (which have proven and proved to be highly uncertain). Furthermore, certain issues that are widely recognized as mistakes in the capital regulations (e.g. inclusion of illusory capital in the capital adequacy ratios) and will ultimately be corrected in Basel III are still allowed in the stress test (transitional rules).
  • Accounting frameworks and EU regulations in different European countries are far from being harmonized. Thus, for a more or less comparable outcome, more or less harmonized data definitions have to be given for the starters. That’s why there are so many templates. (At this stage, however, a number of national idiosyncrasies remain; thus one should carefully read the small texts when comparing banks from different jurisdictions.)
I imagine that this was the task that the EBA’s technical staff was presented: create a building block type of methodology (i.e. a methodology that would enable risk-by-risk analysis of the scenario effects to the banks’ solvency) which integrated banks’ internal pricing and valuation models, risk systems etc. into one common EU-wide framework. Among others, such an approach (pay attention to the underlines):
  • Creates an impression of transparency, this despite that the methodology still heavily relies on the banks’ rather obscure models (read more below).
  • Enables creating an impression of conservativeness (by introducing various caps and floors which in fact may not be all that conservative, for example), even though the amount of debt in the economy and leverage in the financial system as well as feedback effects are not properly taken into account (if at all).
  • Fosters communication between banks and regulators, as well as discussions in the supervisory colleges as there are now common terms and definitions, and a common basis for the talks with the banks.
  • Makes it easier to compare banks, and benchmark their valuations and internal risk assessments; however, the fact that bank supervisors may be able to rank-order banks from the weakest to the strongest (even if in absolute terms, the assessments are misleading) does not necessarily mean that the general public is going to hear the truth – it is still a “black box” of how the conclusions of the Comprehensive Assessment are going to be formed and what role politics will play in it.
Anyways, going now more specific, the EBA has defined five key risks to be paid close(r) attention in the 2014 stress test:
  1. Credit risk – covers all assets in the banking book which are exposed to credit risk excluding counterparty credit risk, on and off‐balance sheet positions;
  2. Market risk – covers all financial assets and liabilities assessed at fair value, including counterparty credit risk; hedge accounting portfolios; securitizations held at fair value;
  3. Sovereign risk (also part of credit and market risks) – covers direct debt exposures as well as indirect exposures to central and local governments; assessed at fair value and amortized cost positions;
  4. Securitization (also part of credit and market risks) – securitization and re-securitization positions assessed at fair value and amortized cost positions; Asset Backed Commercial Paper (ABCP) excluded but subject to either the regular RWA treatment or market risk methodology;
  5. Cost of funding and interest income – covers interest bearing assets and liabilities.
Focus is on the two first risk types (credit risk and market risk). Sovereign risk, securitization and cost of funding are the current issues that market participants are expecting to be highlighted specifically. Then there are small paragraphs for the operational risk, and for non‐interest income and expenses.

For each key risk, the following is described (or prescribed):
  • Terms and definitions,
  • Calculation of the scenario impacts in terms of Profit and Loss (P&L) and Other Comprehensive Income (OCI), and
  • Deriving Risk-weighted Assets (RWAs) for each given risk type.
Capital ratios for each scenario year are then derived as shown in Figure 1:

(Click to enlarge.)

The starting point for the stress test, i.e. the starting capital levels, starting RWAs and starting values of banks’ assets and liabilities as at the end of 2013 are going to be adjusted based on the AQR outcome, though it’s not clear yet how exactly.

I’d draw the reader’s attention to quite a few issues. There are more, but this should be enough for illustrating that the stress test is designed for generating false sense of confidence (not to say: complacency) among market participants and general public. In reality, the ECB will have to find smart ways for removing bad assets from the banks’ balance sheets and arranging distribution of losses.

Firstly, the introduced caps and floors tend to be more favorable for the banks that have been more aggressive in their accounting and risk weightings in the past. For example:
  • Credit RWA is floored at 2013 levels, i.e. the floor is relatively lower for banks that had lower risk weights in 2013. (Recall that internal risk models are not revised during the AQR, thus it’s still so even after adjusting the starting RWAs to the AQR outcome.)
  • For banks using internal models to derive regulatory capital requirements for the trading book, the end of 2013 level of capital charges serve in general as a floor for the [market risk] capital requirements. So the same story as for credit RWA.
  • Capital requirements for operational risk cannot fall below the value as of year‐end 2013 – the above logic is followed here too.
  • By definition, under both scenarios the change in net trading income cannot be larger than zero, i.e. banks cannot have a larger net trading income after stress than at the starting point. In other words: banks that have been more aggressive in their trading activities in times of ultra easy monetary policies are potentially less restricted by this particular cap. 
  • Assumptions underlying the scenarios cannot lead to an increase in net interest income compared with the beginning of the exercise under the baseline and adverse scenario. So the ceiling is among others benefitting banks/countries that are inflating their interest income by accruing interests on non-performing loans. (This accounting practice is in line with the IFRS, but not allowed in all EU countries; also, under the adverse scenario, income on defaulted assets should not be recognized which implies, that the EBA is well aware about the illusionary part in the banks’ reported interest income.)
  • Projected non‐interest income relative to total assets in the base and adverse scenario cannot be larger than the actual 2013 value. In other words: more aggressive valuations of the nonfinancial tangible assets (real estate exposures) as well as valuations of participations are lifting the cap upwards…
It all might be ok, if banks’ risk weightings adequately reflected underlying risks, but if the recent past is any indication, they usually don’t. It might be ok, if the P&L figures for 2009-2013 were not distorted by the banks’ accounting practices and/or by the effects of ultra-easy monetary policies during the past few years, or both; unfortunately P&L figures are distorted.

Secondly, even if a methodology document of 70 pages may seem long, it’s by far not a full stress testing methodology; instead, the described methodology is a mere instruction of how banks should aggregate and present their own assessments based on their own internal methods and models. Among others, participating banks are ought to have the following models in place:
  • Point-in-time PD and LGD models, and satellite models that estimate the relationship between macro‐economic and banking variables for estimating credit impairments (separate models for all portfolio segments, of course);
  • Methodology for the projection of regulatory risk parameters for the calculation of capital adequacy ratios;
  • Internal pricing and risk management models which enable translating the macro‐economic scenarios provided into a stress test impact via gains and losses for positions valued at fair value (applicable to banks with significant trading activities, the so-called VaR banks);
  • Value at Risk models (VaR) and models/methods for estimating the stress VaR (SVaR), incremental risk charge (IRC), comprehensive risk measure (CRM) and own funds requirements for credit valuation adjustments (CVA) in order to calculate market risk capital requirements (applicable to the VaR banks);
  • Methodology for projecting the funding costs and the pass‐through of the change in the cost of funding to the lending rates;
  • Own methodology in projecting non‐interest income and expense paths, i.e. their own perspective on the sensitivity of the respective P&L items to the macro‐economic scenario.
This overreliance on models should make one worried; models should be viewed as tools for providing informational support but not as tools for justifying excessive risk taking and absurdly high leverage in the banking sector which they have become.

Yet there are strong supporters of proprietary models among the regulators. For example, Andreas Dombret who is set to take charge of banking and financial supervision at Germany’s Bundesbank, has recently been reported to say that regulators shouldn’t attempt completely to harmonize the array of models deployed by banks.
“If we unify risk models this could lead to herd behavior, with all those negative effects, and would also exclude the institute-specific measures, which are important for diversity,” he reportedly explained. “I am a big believer in continuing to use risk models...”
Mr. Dombret may have been right about the herd behavior. However, this implies that the stress test conclusions should not be formed based on the regulatory capital ratios but based on simple leverage ratios instead. (Also Mr. Dombret said that models should be counter-checked with leverage ratios.)

Thirdly, certain national accounting differences remain:
  • Provisions for assets that remain as non‐defaulted at the end of the horizon should be computed in line with the accounting systems in each national jurisdiction;
  • Under the baseline scenario, banks are required to project the interest accrued on non-performing loans in line with their standing accounting practice (e.g. no recognition of unpaid income’ i.e. only cash interest received is treated as income, or, full recognition of interest using the original interest rate on the unimpaired balance).
One might argue that in terms of numbers, these differences are not all that significant, but why they are still there?

As for specific risk types, when reading the methodology document, one gets an impression that banks’ market risk positions are just too obscure to be understood by the EBA and the bank supervisors, let alone third persons. Besides heavy reliance on the banks’ internal models, this impression derives from the “Tambov constants”*** in the market risk chapter:
  • Although calculated as the loss resulting from instantaneous shocks to market risk parameters, the impact of the scenario will be distributed over the three years of the exercise (2014, 2015 and 2016), where 50% of loss is allocated to 2014, 30% to 2015 and 20% to 2016, respectively.
  • In the computation of the overall effect on NTI, other comprehensive income or P&L, the net gains resulting from changes within in a major risk factor (e.g. risk factor category “interest rates”) should be reduced by 30% while net losses should be accounted for in full.
  • (Concerning market risk RWA) For correlation trading portfolio, in the adverse scenario the following scaling is assumed to derive the stressed CRM (comprehensive risk measure) capital charge: i. 8 % floor is not binding: 1.5 times the CRM capital charge; ii. 8 % floor is binding: 2 times the floor.
  • … [There are more.]
I wonder why this many caps and floors are needed if supervisors felt comfortable about the banks’ market risk positions.

Yeah, Andrea Enria, chairperson of the EBA, has quoted to say in the press release (stress added): "The exercise's full transparency will be key to its credibility: it will show how efforts recently undertaken by EU banks are already bearing fruit and it will provide a common framework for the next steps to be taken by supervisors and banks."

After reading the methodology document twice and browsing through one more time for my notes, I cannot say that there is or will be all that much transparency for a third person who is not directly involved in the stress testing exercise, this despite of all the communications and publications.

The templates referred to from the methodology probably help supervisors and bank employees to get a better idea about the specifics, but these are not (at least not yet) available for the public scrutiny. Furthermore, transparency data for public disclosure presents a less informative sub-set of the supervisory data. For example, it is explicitly stated that the risk parameters will not be disclosed; detailed funding data is likely to be missing; outsiders will probably have only a vague idea about the effects of the defined caps and floors to the stress test results of individual banks.

I’m repeating myself, but the scenario is probably not the main issue in the 2014 EU-wide bank stress test – the approach for translating it into the conclusions about banks’ solvency may well be. Anyways, I’m planning to explore the baseline and adverse scenarios more closely in my next post in this series.

Meanwhile: Good lack with the AQR and Comprehensive Assessment for those involved! I’m sure it’s not an easy job, no matter what observers like me might say or think about the credibility of the exercise.


____

* Originally, in the 23 October 2013 release, it was communicated that the Comprehensive Assessment would have three pillars or elements: 1) a supervisory risk assessment, 2) the AQR, and 3) the stress test. The supervisory risk assessment which was ought to cover liquidity, leverage and funding of the banks, has been left to the background, as an addition or so (stress added):
“In addition, depending upon the availability of data, a supervisory risk assessment could support the comprehensive assessment through a control/consistency check of the two main pillars in the form of a quantitative and qualitative review of key risks; to include liquidity, leverage and funding.”
It remains a question mark, why the availability of data should be more of an issue for this element than it is for the AQR.

** E.g. Capital Requirements Regulation and Capital Requirements Directive (CRD IV/CRR) are covering more than 400 pages full of gross references, transitional rules etc, and are accompanied by a long list of supervisory reporting templates and hundreds of pages of technical instructions.

*** (Jokingly) Tambov constant is the number which is added or subtracted to the calculation result or with which the result is multiplied or divided in order to obtain the required end result.

1 comment:

  1. Hi, I do believe this is a great web site. I stumbledupon it ;) I am going to return once again since i
    have saved as a favorite it. Money and freedom is the greatest way to change, may you be rich and
    continue to guide others.

    My web blog; stopping foreclosure in Michigan

    ReplyDelete