Systematic or idiosyncratic? Examining the consistency of the probability of default estimates reported by Finnish IRB banks

Abstract

The unjustified variability of risk-weighted assets (RWA) across financial institutions has been a well-known issue since the financial crisis. This challenge is prevalent among financial institutions that are permitted to determine their capital requirement using internal-rating based (IRB) models, as differences in them tend to lead to more or less varying risk parameters for the same counterparties. Academic literature and reports published by regulators show that probability of default (PD) estimates produced by these IRB models often output inconsistent results when compared across institutions.

The empirical part of the thesis examines the consistency of the PD estimates reported by Finnish IRB banks in the Finnish AnaCredit data collection of the Bank of Finland, using a sample from January 2020 to April 2024. The empirical section consists of two parts: firstly, a regression that reveals the associations of a set of counterparty characteristics on the degree of variation in the PD estimates. Secondly, the degree of systematic variation is investigated, i.e. to which extent is the variation systematically explainable by differences in banks’ models.

The results of this thesis are mainly in line with prior literature and regulators’ reports, showing that the systematic variation appears to be of a similar level as in previous studies. Specific to Finnish data, the thesis shows that bank disagreement on the PD estimates of Finnish housing corporations is significantly higher than for other entity types. Considering the large share of housing corporation loans of the entire loan market in Finland, the result suggests that the financial supervisor should look into the PD models regarding these types of entities.