Based on European RMBS deals with 24 million quarterly loan observations, we examine the effect of risk retention on bank behavior. We show that retention deals perform better due to improved monitoring effort and workout processes. We find that the probability of rating changes and collateral revaluations is higher for retention loans, and ratings are more accurate; retention loans have a lower probability of becoming non-performing, a lower delinquency amount, and a shorter time in arrears. Moreover, non-performing and defaulted retention loans are more likely to recover. Reduced losses for deals with retention are associated with lower default rates, lower exposures at default, and higher recovery rates. Our results suggest that retention reduces moral hazard and incentivizes banks to exert higher effort, which results in superior securitized asset performance.

Machine learning (ML) could strengthen banks’ resilience through improved credit risk screening and ultimately benefit financial stability. Yet, ML adoption in banking remains limited, with simpler linear models still predominating. We argue that the emergence of highly flexible ML models has created a new challenge for forecasting tasks: ‘model multiplicity’—where equally accurate ML models at the aggregate level produce divergent individual-level predictions (‘predictive multiplicity’) or differ in their decision surfaces (‘procedural multiplicity’). These issues raise fundamental questions: Why should an individual or firm be subject to an adverse credit risk model outcome when there is an equally accurate model that treats them more favorably? Using the world’s largest loss database of corporate defaults, we examine these two phenomena in recovery rate ( RR ) modeling and propose heterogeneous ML ensembles as a natural solution. By combining predictions and decision surfaces from multiple well-performing ML models, ensembles mitigate risks associated with predictive multiplicity by ensuring that borrowers are not subject to the fluctuations of a single model, and reduce procedural multiplicity by providing a robust measure of features that ultimately improve out-of-sample RR predictions. By addressing the ‘multiplicity of good models’ problem, our study emphasizes the importance of model stability and provides new insights for the future development of ML models.

An accurate forecast of the parameter loss given default (LGD) of loans plays a crucial role for risk-based decision making by banks. We theoretically analyze problems arising when forecasting LGDs of bank loans that lead to inconsistent estimates and a low predictive power. We present several improvements for LGD estimates, considering length-biased sampling, different loan characteristics depending on the type of default end, and different information sets according to the default status. We empirically demonstrate the capability of our proposals based on a data set of 69,985 defaulted bank loans. Our results are not only important for banks, but also for regulators, because neglecting these issues leads to a significant underestimation of capital requirements.

We investigate whether lenders can realize informational synergies by simultaneously obtaining private information from different accounts of the same borrower. Synergies exist if such information is complementary to each other. We focus on consumer credit, using 3.5 million observations from checking accounts and credit card accounts of the same individuals during 2007–2014. First, activity from both accounts is complementary for estimating consumer default beyond credit scores, borrower characteristics and relationship characteristics. Checking accounts display warning indications about consumer default earlier and more accurately than credit card accounts. Second, decision errors are lower when lenders consider cross-product information. The evidence suggests significant informational synergies that are important for the supply and allocation of credit.

Estimating the credit risk parameter exposure at default is important for banks from an internal risk management and a regulatory perspective. Several approaches are common in the literature and in practice. We theoretically and empirically analyze how the exposure at default should be modeled to obtain accurate estimates of the expected loss. Our empirical analysis is based on a large and unique dataset from a retail portfolio of a European bank. We demonstrate that some approaches can lead to substantially biased estimates of the expected loss and show that the generalized cohort approach is advantageous. Moreover, using in- and out-of-sample analyses, we empirically demonstrate that using the credit conversion factor is preferable to the loan equivalent factor, exposure at default factor, and direct exposure at default estimation to achieve high estimation accuracy.