FRS 109’s introduction of the Expected Credit Loss (“ECL”) model has fundamentally changed how provisioning for credit losses is being looked at by banks and other financial institutions. The transition to expected credit loss models requires cross functional support, with expertise around risk management, finance, IT and economic forecasting, being particularly important.

Grant Thornton’s ECL modelling specialists combine such skills and help implement provisioning methodology and process which are right for your business.

Three 'S'

Three ‘S’ define the key success factors for ECL implementation, i.e.

  • Suitability
  • Stability
  • Sustainability
  • Suitability
    Suitability
    Appropriate with reference to size and scale of the business, product profile, customer characteristics, and data/system maturity
  • Stability
    Stability
    Ability to provide reasonable results in various economic scenarios
  • Sustainability
    Sustainability
    Being efficient and accurate, and hence future proof

Design of ECL model

Click on the various components of ECL below to know more about how ECL typically works

Probability of default (PD)

PD refers to the likelihood of a loan defaulting over a time horizon. It is further bifurcated into through the cycle PD (TTC PD) and point in time PD (PIT PD).

TTC PD is calculated as the long run average of historical defaults rates observed by the entity.

TTC PD.png

PIT PD represents forecasted PDs based on current and expected macroeconomic conditions. PIT PD is a function of TTC PD and macroeconomic variables.

Computation of PD is a complex process that involves analysis of data of years of historical experience. We leverage our proficiency in technical accounting and risk management and technological skills to implement flexible and scalable models for our clients.

We have experience of implementing several models for modelling probability of default on varied portfolios. Choice of the model depends on the nature of portfolio and system/data capabilities.

Commonly used methodologies include:

  • Transition matrix
  • Static pool transition rates
  • Markov Chain
  • Flow rates
  • Vintage analysis
  • Survival analysis
  • PD based on credit rating
  • Logistic regression
  • Vasicek model
  • Monte Carlo simulation
  • Discriminant analysis
  • Hazard models
Loss given default (LGD)

LGD is defined as the percentage of exposure that is not expected to be recovered in the event of default.

 

LGD is modelled by analyzing post default recoveries from loans defaulted in the past. Key considerations include considerations of nature and value of collateral, other credit enhancements and cost of recoveries.

Statistical techniques such as regression can be used to incorporate forward looking information in LGD when sufficient historical data is available.

For Stage 3 loans, time elapsed since default date should be considered while applying LGD as per the historical experience of the company.

Our approach on LGD and choice of model adapts to your products and data/ system capabilities.

Generally, the following methods are used to calculate LGD:

  • Workout method
  • Asset pricing models
  • Market LGD
  • Regulatory/ market proxies when sufficient data is not available.
Exposure at default (EAD)

EAD refers to the expected exposure to a borrower in the event of default. The methodology for EAD varies based on the nature of the product.

EAD = Current drawn exposure + (Current undrawn exposure × Credit Conversion Factor)

 

Exposure at the balance sheet date is the key input for EAD.

However, maturity pattern is used to model the expected changes in loan exposure over the lifetime of the asset. Entities are also required to consider the impact of expected prepayments to factor in the behavioral maturity of the asset instead of the contractual maturity.

For revolving facilities and assets with undrawn exposures, credit conversion factor (CCF) is applied to compute EAD. Entities can use workout CCF based on their own data or use regulatory/market proxies. Entities are required to consider their credit risk practices while determining the behavioral maturity for revolving facilities.

 

ECL Computation

ECL reflects an unbiased, probability weighted estimate of cash shortfalls on the loan, and takes into account past events, current conditions and forecasts of future economic conditions.

ECL calculation.png

While PD, LGD and EAD are the key components in ECL computation, the ECL provision is impacted by various other considerations like:

  • Policy on criteria for significant increase in credit risk (SICR) and default definition
  • Policy on individual vs collective assessment and segmentation
  • Overlays: Upon analyzing the output of the model, Management may create additional ECL if the model output does not adequately reflect the credit risk expected in the future.
  • Approach for discounting of ECL

As part of our services, we guide you through navigating the challenges faced while evaluating these considerations to ensure effective ECL implementation.

Our approach to ECL

  • Our approach to ECL

    Policy formation

    The key policies mentioned would be agreed upon in this step.

    The result of this step shall be well deliberated, and documented conclusions on these areas.

    Examples of possible questions to consider include:

    • At what level should segmentation of loan portfolio be done?
    • What should be considered as the threshold of ‘low’ credit risk?
    • What should be definition of significant increase in credit risk (SICR) and default event
  • Our approach to ECL

    Model development

    • Design the methodology for probability of default modelling.
    • Static pool transition matrix is typically done, but in case there is lack of historical data, use of market/proxy data would be necessary. For wholesale portfolio, externally available data like credit ratings is also generally  considered.
    • Identification of macro-economic factors that shall be considered.
    • LGD models for the different portfolios is developed
    • Design custom automated models, where required
  • Our approach to ECL

    Identification and validation of inputs

    • Evaluate the various sources of inputs (both internal and external), that can be used in the models, and advise on the appropriateness of each of them.
    • Evaluate the forward-looking data that should be incorporated.
    • Specify the major estimates and assumptions.
    • Setting up triggers and benchmarks for future update of data
  • Our approach to ECL

    Go live

    • Develop process manual for future reference
    • Document policies for formal internal approval
    • Review and update the models as required
    • Training and presentation to internal and external stakeholders
video banner
The video is playing. This video is playing in mini-player mode.
Chetan Hans
Partner – Head of CFO services
Chetan Hans