PD Modelling

OFS Loan Loss Forecasting and Provisioning PD Modelling accurately predict the number of defaults by omitting optimism or conservatism. Adjustments are made on regulatory capital models to remove inherent conservatism. Estimated PDs are point-in-time. These are adjusted, where necessary, to reflect the effects of the current economic conditions. The PD estimates are recalibrated based on the current representative sample regularly. PD is calculated using sufficient sample size and historical loss data covers at least one full credit cycle.

PD model segments consider drivers in respect of borrower risk, transaction risk, and delinquency status. The PD models are representative of the portfolio segments. The data used for calibration is consistent with the IFRS 9 default definition throughout. Techniques used to determine lifetime PD measures do not introduce bias into the calculation. Future projected macroeconomic factors are used in the computation of lifetime PDs. The risk of default is higher when the expected life of the instrument considering cumulative probability is longer.

The LLFP application calculates Cumulative PD and Final Marginal PD using conditional probability logic. The PD calculation can be performed as mentioned in the following processes:

  • Calculation of Marginal PD whenever Cumulative PD is provided as a download using a Conditional Probability Logic.
  • Computation of the Factor - Lambda, Geometric factor, or Arithmetic factor, depending on the Interpolation technique, using the given formula.
  • Computation of Intermediate Cumulative PD using the given formula.
  • Computation of Intermediate Marginal PD using the conditional probability approach, for each block of time.
  • Computation of Final Cumulative PD by using the conditional probability approach, cumulating across periods till the max maturity bucket. This value is used for the Cash Flow approach.
  • Computation of Final Intermediate PD by subtracting each cumulative PD with the preceding Cumulative PD. This value is used for the Forward Exposure approach.

The whole PD modeling process consists of the following processes:

  • Transition Matrix Definitions
  • Economic Scenarios
  • PD Model Process
  • Survival Model
  • User Group Mapping for Sandbox
  • PD Model Execution Parameters

Note:

Segmentation (Portfolios) is a mandatory prerequisite for these six processes.

The Transition Matrix definition is an optional process. You can also use the transition Matrices generated through the Historical Transition Matrix Run.