The European Central Bank published Working Paper No 3207 setting out a Quantile Probability of Default (QPD) framework to model sectoral corporate default rates across the full distribution, rather than only the conditional mean, to better capture tail risk and heterogeneity in euro area stress testing. The paper notes that the QPD framework has served as the ECB’s benchmark model for assessing non-financial corporate credit risk since 2022, supporting the 2023 and 2025 EU-wide stress tests. The approach uses unconditional quantile regressions with sector and country fixed effects, estimated on sectoral flow default rates constructed from Orbis balance-sheet data for roughly five million non-financial firms across nine euro area countries over 1999–2023, with defaults identified when internal cash flow fails to cover financial expenses in two consecutive years. The model links sectoral default rates to stress-test scenario variables including sectoral gross value added, unemployment, interest rates, term spread, equity prices and property prices, and finds materially stronger sensitivities in the upper tail than around the median, indicating non-linear and asymmetric sectoral propagation of credit risk. Validation exercises show the model tracks crisis spikes and improves out-of-sample projections versus benchmark approaches, and an application to the ECB’s June 2025 severe trade-tension scenario points to higher tail vulnerabilities in construction, wholesale and retail trade, accommodation and food services, real estate and arts and entertainment, with more muted responses in information and communication, transport and utilities.
European Central Bank 2026-03-26
European Central Bank publishes working paper on quantile probability of default modelling for sectoral corporate stress testing
The European Central Bank's Working Paper No 3207 introduces a Quantile Probability of Default framework to model sectoral corporate default rates, enhancing euro area stress testing by capturing tail risk and heterogeneity. Used since 2022, this model supports EU-wide stress tests, showing stronger sensitivities in the upper tail and indicating non-linear credit risk propagation. Validation exercises confirm improved crisis tracking and projections, highlighting higher vulnerabilities in sectors like construction and retail under stress scenarios.