When two banks using the Internal Ratings-Based (IRB) approach merge, the combined institution must consolidate their IRB models and comply with the CRR. A major challenge in such consolidations is how to handle the acquired bank’s historical default data for IRB risk parameter estimation. Often, the acquired institution’s past lending and recovery practices differ from those of the acquirer. In many post-crisis mergers (for example, some Spanish bank integrations after 2010), acquired banks tended to have looser underwriting standards and higher historical default rates than the acquiring bank. This raises a critical question: should the acquiring bank incorporate the acquired entity’s default data into its own IRB models for the calculation of RWA of exposures originated in the acquiring bank before or after the merger, even if the acquired entity’s default data does not reflect the acquiring bank’s current (more conservative) standards? 

Under CRR and EBA guidelines, there are several provisions that frame this issue:

• CRR Article 180(1)(e): Representativeness of PD data: This article mandates that Probability of Default (PD) estimates “reflect current underwriting standards.” If a bank’s underwriting standards have changed or if it acquires a portfolio originated under different standards, any historical default data not aligned with the current standards must be treated with care. Banks are expected to adjust their PD estimates or apply margins of conservatism when using such data to ensure the PD remains relevant to how the bank now operates. This implies that an acquiring bank should not be forced to include an acquired entity’s default data “as is” if that data would make PD estimates unrepresentative of the bank’s actual risk profile after the merger.

• CRR Article 180(1)(h) (and Article 180(2)(e) for retail exposures): Relevant historical observation period: IRB models must use at least 5 years of data (2 years under certain conditions), and if a longer data history is available and relevant, it should be used. The emphasis here is on relevance, meaning the data should be representative of the current portfolio’s risk characteristics. This provides a regulatory basis to exclude or limit older or external data (such as that from an acquired entity) if it is not indicative of the merged bank’s present and future risk environment.

• CRR Articles 181(1)(a) and 182(1)(a): LGD and CCF data requirements: For Loss Given Default (LGD) and Credit Conversion Factor (CCF) estimates, these provisions require institutions to use “all observed defaults within the data sources” for model calibration. In practice, this means all defaults in the chosen relevant dataset must be included, but the bank must first determine which data sources are relevant. If the acquired portfolio’s defaults were deemed not representative and thus excluded from the PD dataset, they should be correspondingly excluded from the LGD/CCF dataset as well. The regulation doesn’t force banks to include extraneous data; it ensures that within the used dataset, no cherry-picking of defaults occurs. Therefore, Articles 181 and 182 don’t override the principle of representativeness, instead, they work in tandem with it. The bank should maintain consistency: the same set of relevant default cases used for PD should feed into LGD and CCF calculations.

• EBA Guidelines on PD and LGD Estimation (EBA/GL/2017/16): Data representativeness and model continuity: The EBA’s guidelines reinforce the CRR’s provisions by explicitly requiring banks to analyze and ensure the representativeness of data used in IRB models. Paragraphs 17–25 of these guidelines list dimensions for assessing representativeness, including lending standards and recovery policies. If an acquired entity’s historical data reflects different lending standards, the guidelines would consider it not representative of the combined bank’s portfolio. In such cases, EBA expects banks to either adjust the data or apply an appropriate margin of conservatism, rather than directly mixing incompatible data. The core message is that IRB risk parameters should remain risk-sensitive and grounded in the bank’s current reality. Using data from another era or institution without adjustments could undermine that risk sensitivity and thus contravene these principles.

In summary, when banks consolidate, they face challenges in their IRB models. The specific scenario addressed by the questions involves an acquired bank’s default dataset that may be inconsistent with the acquiring bank’s risk profile due to different past practices. This has major implications for how the combined bank calculates its capital requirements: the question asks whether the acquired bank’s historical defaults must be included in the merged bank’s IRB PD, LGD, and CCF estimates for the calculation of RWA of newly originated loans and the back-book of the acquiring bank portfolio, even if doing so would violate the “current underwriting standards” criterion. 

This background helps to explain the reasons to raise this question and highlights the relevant regulatory context (CRR Articles 180, 181, 182 and EBA Guidelines) that inform the answer. The overarching concern is to reconcile prudent regulatory compliance with fair and accurate risk measurement after a consolidation.