Stress testing under IFRS 9: a field guide

Higher volatility of loan loss provisioning will complicate financial planning and hit capital

Crash test
Testing times: banks will soon have to calculate stress test results using IFRS 9

Ashutosh Nawani is a director and head of financial risk management and Stanislav Shcheredin a senior manager and credit risk lead at PwC Financial Services Risk Consulting

When it comes into force next year, International Financial Reporting Standard (IFRS) 9 will fundamentally change the way banks calculate expected credit losses for financial assets that are not measured at fair value. As a result, it will dramatically reshape the way dealers conduct statutory stress tests, and in turn lead to a significant increase in the provisions they have to hold against their loan portfolios.

In our experience, provisions under IFRS 9 will typically be more cyclical for banks’ SME and corporate lending business. The difference versus the prior IAS 39 standard could lead to an impact on a bank’s Common Equity Tier 1 (CET1) capital in the range of 180 to 300 basis points, depending on the riskiness of the portfolio and the level of stress under severe but plausible scenarios.

IFRS 9’s approach to recognising impairment is based on a three-stage process. For stage one assets, banks must set aside provisions for 12-month expected credit losses on relevant assets and calculate interest revenue on the gross carrying amount. Stage two applies to assets that suffer a jump in credit risk between reporting dates. Banks must deduct expected losses for these assets’ entire lifetime. Stage three applies to assets that have suffered actual impairments, which must have their interest revenue calculated taking these expected losses into account. 

Measurement of the expected credit losses is determined by a probability-weighted estimate, evidenced by what IFRS 9 calls “supportable information”, including from forward-looking scenarios. This approach represents a sea change from IAS 39, which only forces provisioning once an asset becomes impaired.

As such it poses fresh challenges to stress testing. Banks will frontload the loan losses at the start of a recessionary scenario and, given the likely evolution of such a scenario, will see significant increases in their provisions. This will deplete their loan portfolios’ earning power, impacting capital ratios. The capital base of the bank is eventually going to see an increased CET1 impact due to migration of obligors entering or nearing default.

There are a number of challenges associated with stress-testing IFRS 9 provisions – the difference in the values between the present IAS 39 standard and IFRS 9 being a case in point. Some have tried to estimate the impact, but given impending timelines to implement the point-in-time IFRS 9 provision there has been little published work to understand the volatility of IFRS 9 provisions under stress scenarios once the rules are implemented. One of the important aspects banks will need to understand is how this volatility will impact stress tests and subsequently the CET1 capital ratio.

Higher volatility of loan loss provisions under stress scenarios will complicate medium-term financial planning and threaten the capital adequacy of financial institutions

IFRS 9 stress testing requires financial institutions to revisit their stress-testing capabilities, and pay particular attention to three key factors.

The first is scenario definition and calibration. Unlike stress testing under IAS 39, banks will have to define a number of different scenario paths for each time point in the forecast horizon. Hence, the challenge from scenario calibration is twofold. First, assuming a five-year forecast horizon under regulatory stress testing, banks will have to calibrate a minimum of 15 to 20 additional scenarios under the base case and stress scenario respectively.

Second, probability-weighting of scenarios will have to be made regardless of whether the probability-weighting is assumed to remain constant across the five-year time horizon or changes in line with banks’ own assumptions on how the weights on various scenario paths may change as the economy dips into a crisis and subsequently recovers in the outer years.

The second factor concerns the forward-looking assessment of provisions. The extension of the IFRS 9 models for stress testing will complicate assumptions for transition criteria under stress scenarios, stressed lifetime probabilities of default, stressed behavioural assumptions, and discounting, among others.

The three-stage process for estimating provisions poses a further challenge. The migration of assets between the three stages (stage 1: point-in-time expected loss; stage 2: lifetime expected loss; and stage 3: defaults) is a function of quantitative risk measures as well as individual banks’ own policies on what is considered a significant deterioration in loan quality. Uncertainties in estimations can result from the definition of migration from stage 1 to stage 2. Qualitative and behavioural factors will need to be factored in to ensure the stress-testing results bear semblance to what the bank observes on its balance sheet as reported provisions.

The third factor concerns business assumptions. Further uncertainties in IFRS 9 provision stress testing will result from key business assumptions that underpin the financial and business strategy of the bank. These include assumptions around portfolio movements, product mix, underwriting criteria, product maturity, limit management, interest rates across the product mix, fee income estimates, and so on.

The uncertainties inherent in these three factors will translate into higher volatility of loan loss provisions under stress scenarios. These in turn will complicate medium-term financial planning and threaten the capital adequacy of financial institutions.

Quantifying the impact

Below, we estimate the volatility of provision calculations for a representative SME and corporate lending bank portfolio, subjected to five years of stress testing.

A typical stress-testing exercise defines two deterministic scenarios for key macroeconomic risk factors: a base scenario and a stress scenario. Factors such as GDP, unemployment, house prices and interest rates are taken into consideration. Examples of narratives for such economic stresses can be a sudden drop in house prices, a slowdown in emerging markets with a spillover effect on developed economies, or conduct related penalties on major tier one banks resulting in a weakening of their financial positions and knock-on effects on lending to the real economy.

IFRS 9 chart

After making a number of assumptions around the calculation of the stressed provision values, including the assumption that loans defaulted under stress do not recover, we obtain the following results for distribution of assets by stages:

IFRS 9 chart

The forwards and backwards transition between stages leads to a high volatility in the provision values as illustrated by the below figures.

IFRS 9 chart
IFRS 9 chart

Clearly, the above results are subject to a number of the aforementioned uncertainties. For example uncertainty in the choice of IFRS 9 scenarios can lead to a minimum 4% uncertainty in the above results. We estimate that overall the uncertainty about the above figures can be as large as 10%.

The above analysis shows increased procyclicality in the provision estimates, in particular under stressed market conditions. The impact is going to be amplified further if banks do not pay close attention to their choice of methodology, its assumptions and limitations, its evolution through scenarios, and distributional choices. Nor can qualitative assumptions around originations, forbearance or behavioural maturities be discounted.

As seen in the case of traditional stress-testing modelling, model limitations have eventually led to analysts and senior management challenging the uncertainties around them. Such interrogations will become more common under IFRS 9 as banks seek to justify their chosen assumptions and methodologies.

The views expressed in this report are the authors’ alone and do not necessarily represent the views of PwC or affiliated firms.

  • LinkedIn  
  • Save this article
  • Print this page  

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an indvidual account here: