Automated Valuation Modeling

Selecting the right Current Expected Credit Loss (CECL) model is a critical decision for every institution, and it must be done uniquely for each pool of assets. The process requires careful consideration of historical data, including charge-offs, defaults, and the presence of outliers, as these factors significantly influence a model's predictive power. This ARCSys whitepaper provides a comprehensive overview of the models utilized by ARCSys, including the Discounted Cash Flow (DCF), Probability of Default (PD), and Open Commitments models. We explain how each calculation works and outline a robust, six-step model selection process—from initial questionnaire and expected allowance estimation to final statistical evaluation using metrics like Root Mean Square Error (RMSE) and Relative Root Mean Square Error (RRMSE).

For institutions navigating the complexities of CECL, this guide is an essential resource for ensuring your model results are reasonable, supportable, and a "good fit" for purpose. It details the required due diligence, including the critical steps of reviewing forecast charts for reasonableness and utilizing statistical tools to check for model accuracy and prevent over- or under-fitting. Ultimately, model selection demands informed judgment, supported by thorough documentation of segmentation, forecast reviews, and statistical analysis. Let ARCSys assist your institution with expert modeling decisions, documentation, and advanced analyses like Back-Testing and Benchmarking to fully support your CECL calculations.