A fresh take on risk and valuation
Back
Article

2018 stress tests: evolution, revolution or a chimera?

By Jean-Pascal Kretz, Senior Consultant

and  Kristian Lajkep, Regulatory Compliance Officer

Outline of Article

Introduction

By 2 November 2018, the European Banking Authority (EBA) will conclude and publish the results of the 2018 stress test.

Stress tests have been increasingly relied upon since 2008, as regulatory bodies are looking for insights on the ability of the financial industry to resist to extreme, negative evolutions of their economic and financial environments from P&L and regulatory capital perspectives.

The primary focus is the assessment of the impact of the following risk drivers on banks’ solvency:

  • Credit risk, including securitisations
  • Market risk, Counterparty Credit Risk (CCR) and Credit Valuation Adjustment (CVA)
  • Net Interest Income (NII)
  • Operational risk, with an extensive focus on conduct risk

Additionally, projections of the effect of the scenarios on Non-Interest Income and stress to Profit & Loss statements and capital items not covered by other risk types are expected as well.

Led by the EBA in cooperation with the ECB and the European Systemic Risk Board, the 2018 stress test has been conducted on about 70% of the total consolidated assets of banks from the European Union and Norway by addressing the 48 largest institutions1 in the EU and Norway. Moreover, non-participating banks are expected to run the exercise on a voluntary basis.

1 - 49 institutions initially, down to 48 after the Bankia/ Bank Mare Nostrum merge

Its results will provide national authorities with a unified and consistent base of comparison of the resiliency and ability to react of the banking industry. The impact of the EU-wide stress test will be reported in terms of CET1 capital, Tier 1 and total capital ratios and leverage ratio, for every year of the exercise.

Following suit with the 2016 exercise, there is no set pass/fail threshold. Instead, the exercise
ties in with the Supervisory Review and Evaluation Process (SREP), with significant findings
used as inputs to the relevant SREP at the individual bank’s level.

As regards publications, the EBA will provide outputs similar to the previous exercises, with bank-level data, aggregate reports and interactive tools so that all market participants are able to determine how banks are dealing with remaining pockets of vulnerability.

It is worth noting the potentially negative impact of such a static approach, by definition negating the possibility of a dynamic portfolio risk management and asset substitution for better creditworthiness or higher liquidity.

A 2018 Stress Test similar to the 2016 exercise in its overall approach:

 Common approaches between 2016 and 2018
Overall
Approach
Bottom-up projection from banks under constraints, considering a static balance sheet
Static balance sheet assumptions• Assets and liabilities maturing within the time horizon of the exercise are replaced with similar financial instruments in terms of type, currency, credit quality at date of maturity, and original maturity.

• Banks maintain the same business mix and model throughout the time horizon.

• If the euro is not the reporting currency, all stock projections should be translated by applying the exchange rate as of 31 December 2017.

• Static assumptions apply to institutions subject to mandatory restructuring plans formally agreed with the European Commission if included in the sample.
Pass / Fail IndicatorNo set pass / fail indicator will be communicated. The 2018 stress test will result in three years of leverage ratio, CET1 and total capital projections.

The exercise is meant as an input for the Supervisory Review and Evaluation Process (SREP) between bank and supervisor.

The introduction of IFRS 9 could constitute a game changer… or could have

The IFRS 9 expected credit loss framework forces banks to assess their assets in terms of stages, based on the payment behaviour and the evolution of the credit risk of the counterparty and not only in terms of flow from performing to defaulted.

In effect, the following expected credit losses are calculated:

  • A minimal, non-zero expected credit loss for performing assets with a good credit standing and positive outlook, calculated on a one-year horizon (stage 1 assets)
  • Switching to a lifetime expected credit loss on the asset if the counterparty is deemed to be undergoing a “significant increase in credit risk” or if the payment of the financial obligation is overdue by more than 30 days but still performing (stage 2 assets)
  • If the counterparty effectively defaults, the expected credit loss remains calculated on a lifetime basis and the asset is allocated to stage 3, an absorbing state in the 2018 stress test

Therefore it is generally expected that IFRS 9 will impact the equity capital ratios negatively:

  • by creating a permanent increase in loan loss provisions with the introduction of the stage 2 exposures and
  • by the sensitivity of the forward-looking default probability approach to negative macroeconomic factors, which will substantially increase the volumes of assets in stage 2 during rougher times and thus loan loss provisions, reducing capital further.

Interestingly enough and conversely to the IFRS 9 standards and the Basel Committee guidelines, the stress test is quantifying this significant increase triggering the allocation to stage 2 of the exposure as a “threefold increase of lifetime PD with regard to lifetime PD at initial recognition”, which is a de facto limit to the discretion level banks could apply.

However, despite the certainty that IFRS 9 will have an important impact on the stress tests in the future, the introduction of the European transitional arrangements will in fact allow banks to fully neutralize any negative impact in the 2018 stress test. Even worse, certain provisions that would exist under IAS 39 might be subject to the transitional arrangements and therefore added back to CET1. This would in effect increase the CET1 for certain banks, effectively having fully unintended results1.

1 - See the opinion of the EBA here

A closer look at the macroeconomic scenarios at play

The Stress Test makes use of a baseline and an adverse scenario. The baseline scenario reflects and accounts for the average of a range of possible outcomes of UK leaving the European Union (Brexit), whilst many possible risk factors that may arise as a result of Brexit are captured in the adverse scenario.

The adverse scenario is the realization of four systemic risks, as stated in the ESRB “Adverse macro- financial scenario for the 2018 EU-wide banking sector stress test”:

  • Abrupt and sizeable repricing of risk premia in global financial markets, leading to a tightening of financial conditions
  • Adverse feedback loop between weak bank profitability and low nominal growth, amid structural challenges in the EU banking sector causing a credit crunch
  • Public and private debt sustainability concerns amid a potential repricing of risk premia and an increased political fragmentation
  • Liquidity risks in the non-bank financial sector with potential spillovers to the broader financial system.

On the first two items selected, the 2018 exercise is visibly significantly more severe than the same scenario under the 2016 EBA stress testing exercise. As regards the adverse unemployment rate, though the impact seems higher under the 2016 scenario, the baseline unemployment rates are largely improved in 2018 – from baseline unemployment rates of 9.2%, 8.9% and 8.9% for 2016 - 2018 to 7.1%, 6.7% and 6.4% for 2018 - 2020.

Broadly speaking, in comparison with 2016, the adverse scenario of the 2018 stress test exercise is much more severe. The deviation of the EU GDP from the baseline under the adverse scenario is greater than before. Also, additional domestic real and financial shocks have been incorporated for countries enjoying rather better macroeconomic factors for the purpose of comparability of results, making the scenario even more severe in those countries – although the mechanisms and impacts thereof can be disputable.

Still, as evidenced in the in-depth analysis of the consistency and efficiency of the 2018 stress test1, it is likely that the 2018 stress test scenarios translate into lower credit losses than in 2016 ceteris paribus. An explanation is that the unemployment rate has a large weight in the 2018 methodology. A second explanation is that the 2018 scenario tends to have less effect on banks that were especially affected by the 2016 scenario.

1 - See The European Parliament INDA

Relative severity of the adverse scenario in the 2016 and 2018 exercises on selected items:

 2018 ST 2016 ST Cumulative Impact
 2018201920202016201720182018 ST 2016 ST
GDP Growth
(EU countries)
-1,20% -2,20%0,70% -1,20%-1,30%0,70%-2,70%-1,80%
Residential Real Estate Prices
(EU countries)
-9,60%-9,80%-0,80% -7,70%-2,90%-0,60%-19,11%-10,91%
Unemployment Rate
(EU countries)
7,90% 9,00%9,70% 9,90% 10,80% 11,60% 29,02% 35,89%

 

Detailed measurements: Credit Risk

The credit risk methodology for the 2018 exercise is unchanged. The method includes a prescribed increase in the risk-weighted exposure amount (REA) for securitisation exposures, as well as prescribed shocks to credit risk losses for sovereign exposures.

As previously, the securitisation positions not in the correlation trading portfolio are covered under the credit risk methodology.

Here is an overview of the changes introduced in the 2018 stress test, as well as the key assumptions and constraints:

 New in 2018 Assumptions taken Major constraints applied
Credit Risk Methodology• Transitions of exposure between the three impairment stages defined in IFRS 9 need to be projected for each year.

• Banks are expected to provide stressed lifetime expected loss rates for exposures in Stages 2 and 3.

• The EBA stress test explicitly defines the "significant increase in credit risk" (SICR), not addressed by IFRS 9 or the Basel Committee previously.
• Perfect foresight approach: the full scenarios are considered as known when calculating expected credit losses.

• The projection of provisions is based on a single scenario in each macroeconomic scenario.

• The baseline credit risk parameters are assumed to stay flat after year 3. This will largely impact the transition from stage 1 to stage 2 for long term assets.

• For the estimation of lifetime ECL, adverse scenario credit risk parameters revert to the baseline horizon credit risk  parameters after the end of the scenario horizon.

• A common definition of Stage 3 assets as non-performing exposures should be applied for the projections.
• No negative impairments, except and exclusively in the case of transitions from stage 2 to stage 1; i.e. after its allocation to stage 2, SICR back below the threshold and exposure not past due for more than 30 days.

• The coverage ratio for stage 1 assets cannot decrease.

• No cures for stage 3 assets (non performing exposures), i.e. no transitions from stage 3 to stage 2 or 1.

 

Detailed measurements: Market risk, CCR losses and CVA

As was the case for the previous exercise, Market Risk is calculated based on a set of stressed market parameters, calibrated on the macroeconomic adverse scenario.

Some differences, and the attached assumptions and constraints, are developed below.

 New in 2018 Assumptions taken Major constraints applied
Market risk, Counterparty Credit Risk (CCR) and Credit Valuation Adjustment (CVA)• Impact to be assessed through a full revaluation of all asset categories (comprehensive approach) with fair value measurement under IFRS 9 or all assets and liabilities (trading exemption approach) with a full or partial fair value behaviour except for the items held with a trading intent and their related hedges.

Valuation reserve for level 2 and level 3 instruments will be stressed to account for modelling uncertainties related to these instruments.

The impact of FX risk on the banking book and related hedges is excluded; under the trading exemption, banks are allowed to apply related hedges to HFT items.

Furthermore, consistently with the full revaluation, banks have to recalculate the credit valuation adjustment (CVA) and liquidity reserve based on the adverse macroeconomic scenario.
• No deviation from the starting value under the baseline scenario.

• Full revaluation impact on HFT items and their related hedges capped at a haircut of the sum of asset and liabilities under the adverse scenario.

• The baseline value of not trading income (NTI) is the minimum of the historical value averages across the last 2, 3 and 5 years (2-year average floored at 0).

• Under the adverse scenario, client revenues projections are capped at 75% of 2017 annual client revenues and 75% of the baseline NTL.

• REA constant in the baseline scenario, cannot go below its starting value in the adverse scenario.

• REA a multiple of the risk measures for VaR and all price risk (APR).

• Banks without an approved VaR model maintain market risk REA constant at the starting value for both the baseline and adverse scenarios.

• Impact on REA for incremental risk charge (IRC) and CVA floored at the increase for IRB REA.
• No negative impairments, except and exclusively in the case of transitions from stage 2 to stage 1; i.e. after its allocation to stage 2, SICR back below the threshold and exposure not past due for more than 30 days.

• The coverage ratio for stage 1 assets cannot decrease.

• No cures for stage 3 assets (non performing exposures), i.e. no transitions from stage 3 to stage 2 or 1.

 

Detailed measurements: Net interest income (NII)

From a methodological perspective, NII largely retains the principles for the 2016 exercise, elaborating on IFRS 9 standards with the inclusion of the effective interest rate (EIR).

Banks may use their own methodologies and earnings at risk models to project their net interest income on all interest-earning or interest-paying positions across all accounting categories, under either the amortised cost or fair value measurement.

Banks may rely on their own assumptions regarding the pace of the repricing of their portfolio, and their projections for risk-free reference rates and margins under both the baseline scenario and the adverse scenario.

Finally, the reference rate and margin components of banks’ assets and liabilities are treated separately in order to distinguish risks affecting banks’ NII under stress, namely a change in the general ‘risk-free’ yield curves and the risk related to a change in the risk premium.

Banks’ projections are subject to the following constraints:

  • Under the adverse scenario, assumptions cannot lead to an increase in the bank’s NII compared to the value indicated for 2017.
  • Under the adverse scenario, banks are required to project the income on non-performing exposures net of provisions, subject to a cap on the applicable EIR.
  • Under the baseline scenario, banks are required at a minimum to reflect a proportion of the changes in the sovereign bond spread of the country of exposure in the margin component of the EIR of their repriced liabilities.
  • Under the adverse scenario, the margin paid on interest-bearing liabilities cannot increase less than either a proportion of the increase in the sovereign spreads of the country of exposure, or the proportion applied to the increase of an idiosyncratic component, derived from the impact on banks’ wholesale funding rate of a rating downgrade; whichever is the higher.
  • Banks are required to cap the margin component of the EIR on their repriced assets at a proportion of the increase in the sovereign spreads of the country of exposure.
  • It is expected that the change in the reference rate of new originated or repriced instruments is consistent with the macrofinancial scenarios for risk-free yield curves.

It therefore appears that the net interest income is capped by construction, given that it must not
exceed the current income levels under the adverse scenario.

Detailed measurements: Conduct risk and other operational risks

“Conduct risk”, under which increasingly large losses have been reported in recent years, is now receiving a particular focus.

Banks are expected to use all available qualitative information to do their own forecasts for these losses. This forecast is nevertheless subject to floors set by the authorities which are based on the individual banks’ historical loss experience.

The following constraints apply:

  • Projections of losses from new non-material conduct risk events are subject to a minimum overall 3-year floor, calculated as:

 

         and

 

          where year(i) is in the period 2013 - 2017.

  • Projections of conduct losses connected to material conduct risk events are subject to a floor in the quality assurance process, i.e. banks that submit projections which are lower than the floor are required to justify their projections to their competent authority
  • Projections of losses due to other operational risks are subject to floors as well, calculated as follows:

 

         and

 

         where year(i) is in the period 2013 - 2017

  • The total capital requirements for operational risk in each year of the projection horizon shall not fall below the actual minimum capital requirements for operational risk reported by the bank on 31 December 2017
  • In the absence of relevant historical losses and/or projections, overall operational risk loss projections, aggregated for the 3 years of the exercise, will be calculated as a function of the relevant indicator (6% of the RI and 15% respectively in the baseline and adverse scenarios)
  • For the operational risk categories where the capital requirements are calculated using basic and standard approaches, the capital requirements shall stay constant and equal to the capital requirements reported by the bank for the starting point (31 December 2017)

Conversely to the previous exercise, the 2018 stress test enforces a floor for both material and non-material conduct losses. Nevertheless, by providing a “reasonable” justification to its  competent authority, a bank may deviate and use more optimistic projections for material conduct loss.

Banks shall project the Profit and Loss impact reflecting the losses arising from conduct risk and other operational risks, using, when relevant, their internal models. Surprisingly, despite the Basel Committee on Banking Supervision requesting a phase out of the advanced measurement approach (AMA) in favour of the standardized measurement approach (SMA) for operational risk over the coming years, the 2018 stress test methodology only focuses on the AMA.

Finally, the 2018 stress test still applies the fall-back solution for cases when banks cannot provide historical loss data, with losses being calculated as a function of gross earnings. However, this fall-back solution effectively lessens information value, since it has massively been used by the participants in 2016. This is not a strong incentive for banks to invest in tools for the proper estimation and management of operational risk, despite it being increasingly important.

Non-interest income, expenses and capital

Applicable to P&L and capital items not in scope of risk types or NII, the relevant projections prescribe that the macroeconomic shocks and market risk methodologies shall be applied for stressing real estate assets and defined benefit pension plans, respectively.

Applicable constraints are as follows:

  • Prescribed caps for dividend income, NFCI and share of the profit of investments in subsidiaries, joint ventures and associates
  • Floor/cap for administrative expenses, depreciation and other provisions or reversal of provisions, other operating income and expenses
  • Limitation of the number of one-off adjustments and permitted as well as excluded cases
  • Prescribed threshold for recognition of submitted one-off adjustments
  • Prescribed floor for dividend payments and link between the baseline and adverse scenarios
  • Prescribed approach for distribution restrictions under Article 141(3) of the CRD
  • Application of the common tax rate
  • The creation of DTAs that rely on future profitability and do not arise from temporary differences is limited to the offsetting of negative pre-tax profits
  • Prescribed floor for DTAs that do not rely on future profitability
  • No impact for realised gains or losses, gains or losses on derecognition of non-financial assets, modification gains or losses, negative goodwill, impairments on goodwill, foreign exchange effects
  • Inclusion of adjustments on capital items for the new implementation of IFRS 9 as of 1 January 2018
  • Prescribed formula for transitional adjustments to CET1 capital from unrealised gains/losses

Despite quite frequent prescriptions, one of the few cases where the stress test methodology is changed in a way that allows banks to exercise more discretion and thus reduces the severity of the stress test can be found here.

Indeed, the projection of dividend income from the subsidiaries of a bank is no longer prescribed, as it was in the 2016 stress test.

In the 2018 exercise, banks are allowed to use their own projections, which could be more favourable. Even though these projections are subject to a floor, this floor is less conservative than the 2016 stress test rules, with the 2018 floor for the adverse scenario based on current dividends, while the 2016 stress test required to use the minimum of current dividends and the average of the two smallest dividends in the last 5 years.

Conducting the stress test

It is crucial that:

  • Methodology, definitions and assumptions are well-documented.
  • Results are tractable, robust and make sense empirically.
  • The regulatory changes applicable are well understood, in particular IFRS 9.

The stress test can be divided into the following steps:

  • Identification of the scenarios – while the EBA is providing a detailed description of the scenarios necessary for its stress test, many institutions want to take advantage of already having to run stress tests in order to do stress testing for other contingencies for various different purposes – IFRS 9 being one of these, despite obvious differences in the stance taken; plausible scenarios for IFRS 9 as opposed to extreme scenarios for stress testing for instance.
  • Translation of the above scenarios into industry-relevant scenarios.
  • Design of the stress test model: the framework that helps to gain insight in the impact the scenarios have on the portfolio value in question (e.g. defined scope, data sources, definitions/assumptions, etc.).
  • Impact analysis: Gain insight in to what extent the factors in the model can explain the change in the portfolio value due to the scenario (e.g. expert opinions, in case of a qualitative approach).

Finalyse is experienced in all these matters, including model validation and IFRS 9 specific solutions.