Loan Recovery Determinants -- A Pan-European Study

Tail approximations in credit portfolios using large deviations techniques

“…Some studies suggest higher values for the expected RR and especially for the standard deviation of the RR (see e.g. [24]). Let the setting be the same as before, only RRs have a different distribution, with a higher expected RR and a higher standard deviation of the RR.…”

“…In several studies, it is found that there is a negative correlation between default and RRs (see e.g. [18], [2], [3] and [24]). If the economy is fine, default probabilities tend to be lower with on average higher RRs.…”

“…Several studies state a negative correlation between default and RRs (see e.g. [3], [8] and [24]). In the context of banking supervision, some experts already suggest to estimate loss given default, defined as LGD=1-RR, depending on the state of the economy, to better account for recovery risk (see [4]).…”

Cyclicality in losses on bank loans

“…If, on the other hand, the full debt is recovered, including penalty fees, legal costs, and principal advances, the amount received during recovery exceeds the EAD and the LGD is negative. In line with Höcht and Zagst (2007) and Hartmann-Wendels et al (2014), LGD observations below −0.5 or above 1.5 have been removed.…”

“…NIBC's LGD dataset contains 92,797 loans with 46,628 counterparties We exclude nonrepresentative observations based on Höcht and Zagst () and NIBC's internal policy (see Supporting Information Appendix B.2 for details). Following NIBC's practice, we discount all cash flows by the 2‐year swap rate plus the spread from the loan.…”