This paper sets out a methodology for estimating the impact on a portfolio’s expected credit loss (ECL) caused by potential model risks in underwriting, such as those arising from scoring models (for retail banking) and rating models (for wholesale banking). Such models are used not only for underwriting assessments but also for the determination of through-the-cycle probabilities of default (for regulatory capital purposes) and point-in-time probabilities of default (for International Financial Reporting Standard 9 ECL reporting). An important metric in credit risk management and assessment – required by regulators and auditors – is the impact on ECL if the model’s Gini coefficient deteriorates. We establish the quantum of Gini weakening that breaches a bank’s internal validation standards (as assessed by a consequently large impact on the ECL), and we formulate an analytical approach, caveated with realistic assumptions, that allows for tractable, computationally simple measurement of ECL as the Gini coefficient worsens. The efficiency gained by the simplifying assumptions comes at the cost of losing granularity and some degree of accuracy.