Riskworx was approach by a financing company to review their model for calculating ECL (Expected Credit Loss) for the purpose of IFRS9 Credit Impairment reporting.

Solution:
The existing model was found to be very Euro-centric, leveraging information from an international credit agency that did not distinguish between developed  and developing countries in its calculation of the Probability of Default (PD), leading to inaccurate ECL figures. Riskworx developed a model where the lifetime PD is calculated using a combination of statistical methods (including multiple and logistic regression) and business logic.  This lifetime PD was converted into a point-in-time (PiT) PD that evaluates the probability of default at any future point in time. It considers both macro-economic factors and the risk attributes of the borrower.  Since it considers future changes in macro-economic conditions, the PiT PD moves up as macro-economic conditions deteriorate and moves down as these conditions improve. Riskworx used a neural network model to forecast changes based on historical observations.

“Riskworx used a neural network model to forecast changes based on historical observations.”

Value
Enhanced PiT PD as well as more accurate ECL figures. The client was able to improve customer credit risk management as well as loan origination.