A project that can only be described as a demanding quantitative project that required ingenious approaches; a South African-based Developmental funding institution tasked Aspect Advisory with validating its existing risk-based pricing tool and expected credit loss (ECL) rating tool and methodologies as deployed within the risk management framework; along with identifying areas within the pricing and rating tools that might require recalibration. The project also included assessment of the organisation’s pricing framework to identify key drivers that influence loan pricing.
Aspect Advisory employed a custom-made validation framework suited for the DFI’s mandate and project requirements. We broke down the validation and review exercise into three distinct components: qualitative validation which evaluated the design philosophy in accordance with Basel Accord and other supplementary guidelines, quantitative validation which incorporated statistical tests to assess models’ performance and identify areas for recalibration, and finally, benchmarking of model application with peer financial institutions.
Validation of Expected credit loss (ECL) models showed the probability of default (PD) model to have less predictive power as compared to model inception which warranted a complete recalibration of the PD model starting with correlation assessment of financial factors; calibration of LGD model was done by utilising the Basel benchmarks for unsecured exposures (75%) and the minimum LGD (35%) typically used for mortgage exposures for a level of overcollateralisation of 140% implying that the model did not incorporate historical observations in its calibration; the EAD model was found to be fit for purpose. The pricing tool itself proved to be insensitive to interest moratorium and insensitive to exposure, aspects that were reconfigured to best practices by Aspect Advisory.
Risk management, Treasury, Quantitative analysis and Validation
Risk-based pricing is becoming increasingly popular with financial institutions. The ECL models make up an integral part of this framework thus the model’s reliability and precision are to be constantly monitored. It is up to the bank’s risk management team to stipulate validation and recalibration frequency, with a recommendation of once every year. In this existing project, we discovered that first-generation models are dependent on Basel guidelines and need to be recalibrated as soon as a rich set of historical data has been collected. In this regard, data collection and storage should be at the forefront of model ownership and usage with dedicated frameworks being assigned to this.