Credit Loss & Provision Modelling
This use case explains how to model expected credit losses, charge offs and provisions for loan portfolios in Model Reef.
You will:
Attach expected loss assumptions to loan products and segments.
Model charge offs, recoveries and provision movements over time.
Link credit losses to P&L, Balance Sheet and Cashflow.
Use scenarios to test credit quality under different macro conditions.
Model Reef is not a full IFRS 9 or CECL engine and does not replace regulatory tools. It provides planning level credit loss and provision modelling aligned with loan book forecasting.
When to use this pattern
Use this pattern when:
Credit losses are material to your lending economics.
You want to connect loan book growth with expected loss behaviour.
You need scenario views of credit quality, losses and provisions.
You want high level but transparent ECL style modelling.
It builds on:
Loan Book Growth Forecasting
Multi Product Financial Model
Build a Stress Test or Downside Case
Build a Multi Scenario Valuation Pack
Architecture overview
Credit loss and provision modelling uses:
Segmentation
Portfolios and products.
Risk bands or credit score ranges.
Vintage or cohort if you want more detail.
Loss drivers
Probability of Default (PD).
Loss Given Default (LGD).
Exposure at Default (EAD).
Recovery rates and timing.
Accounting representation
Expected credit loss expense in P&L.
Provisions on the Balance Sheet.
Charge offs, recoveries and net write offs.
Scenario logic
Different credit cycles and macro conditions.
Portfolio mix and underwriting changes.
All of this is implemented via driver based variables in the model, not procedural code.
Define credit segments and attach to products
Decide how you want to segment credit risk, for example by:
Product (home loans, SME, consumer).
Risk band (for example internal rating grades or credit score ranges).
Geography or channel if behaviour differs meaningfully.
In the branch tree or in variable naming and categories, ensure you can identify:
Loan exposures for each credit segment.
Associated interest and fee income.
This will allow you to apply segment specific PD, LGD and EAD drivers.
Build PD, LGD and EAD drivers
In the Data Library, create drivers for each segment such as:
PD per period or annualised PD.
LGD as a percentage of EAD.
EAD multipliers if on balance sheet exposures differ from contractual limits.
Recovery rates and lag structures where you approximate recovery timing.
You can base these on:
Historical performance.
Internal risk models.
External benchmarks and expert judgement.
For simplicity at planning level, many users assume:
Point in time or through the cycle PD paths.
Relatively stable LGD per segment, adjusted for scenarios.
Compute expected credit losses and provisions
Create variables for expected credit loss for each segment, for example:
ECL - Home Loans - Prime.
ECL - SME - Standard.
Conceptually, expected credit loss per period can be approximated as:
ECL Expense = PD × LGD × EAD.
Where EAD can be proxied by period start or average balances, adjusted by EAD multipliers for undrawn exposures if you include them.
Represent provisions as Balance Sheet variables that roll as:
Closing Provision = Opening Provision plus ECL Expense minus Net Write Offs.
You can also distinguish between:
Stage based or life time losses if you want to mimic IFRS 9 at a high level by varying PD and LGD profiles across horizon, but Model Reef does not enforce a particular accounting standard.
Model charge offs, recoveries and net write offs
Define drivers for:
Charge Off Rate for bad debts per segment.
Recovery Rate on charged off amounts.
Recovery Delay in periods.
Use these to compute:
Charge Offs = Charge Off Rate × Relevant Balance or EAD.
Recoveries = Recovery Rate × Past Charge Offs shifted by recovery delay.
Net Write Offs = Charge Offs minus Recoveries.
Link net write offs to the provision roll above. This will produce:
Provision movements in the Balance Sheet.
ECL expense in P&L.
Cashflow impacts for loss and recovery timing in the Cashflow Statement.
Integrate with loan book growth and funding models
Because Loan Book Growth Forecasting already produces balances and interest income, you can:
Use those balances as EAD approximations for credit loss calculations.
Combine interest income with credit loss expense to derive risk adjusted yield.
Analyse net interest income after credit losses (NIM after losses).
In group reporting and valuation, credit losses will:
Reduce P&L through ECL expense and net write offs.
Reduce Balance Sheet assets where losses are realised.
Reduce free cashflow and hence valuation.
This ensures that growth, risk and value are all consistent in one model.
Use scenarios to test credit cycles and underwriting changes
Clone the base model into scenario models to explore:
Different credit environments (benign, normal, stressed).
Changes in underwriting standards and portfolio mix.
Economic downturns with higher PD and LGD.
Improved collections and recovery performance.
In each scenario, adjust:
PD, LGD and LGD paths over time.
Charge off and recovery rates.
Mix of exposures across segments and products.
Loan book growth and pricing assumptions for consistency.
Compare scenarios using:
Loss rates and provision levels.
Net interest after losses.
Capital and funding implications where you approximate them.
Valuation impacts via changes in FCFF and FCFE.
Check your work
Loss and provision levels are in line with historical experience for base scenarios.
Stress scenarios are severe but plausible and documented.
Losses scale sensibly with loan book growth and mix.
Stakeholders understand the simplifications relative to formal regulatory models.
Troubleshooting
Related guides
Last updated