According to the Basel II accords, models used for risk estimation need to be validated on an annual basis. In order to validate credit risk models, we comprehensively review the structure, calibration and operation of internal models and systems. We align the validation exercise to four key modules of model validation: Preliminary model review, Qualitative validation, Data preparation and analysis, and Quantitative validation. At each module, we leverage our proprietary accelerators to fast track the process. 

Some of the key activities that we perform during the exercise are as follow:

  • We analyze the current structure and methodologies used by the bank and benchmark them against Basel standards and industry best practices
  • We analyze the effectiveness of the models, perform sensitivity analysis of variables within the models and recommend changes to enhance effectiveness
  • We use various statistical techniques to evaluate the discriminatory power of rating models, including
  • Cumulative Accuracy Profile (CAP) methodology, and its summary Accuracy Ratio
  • Information Value
  • K-S Test
  • We perform individual factor (variable) level analysis to suggest scope of model performance enhancement at the ground level
  • We test the efficiency of model calibration with techniques like
    • Binomial Test
    • Normal Test
    • Chi – Square Tests
    • Traffic Lights Approach

       Our approach towards LGD and EAD model validation are in line with leading market practice and Basel regulatory requirements