Banks' credit loss forecasts: lessons from supervisory data

BIS Working Papers  |  No 1125  | 
25 September 2023



After the Great Financial Crisis, policy initiatives sought to overhaul banks' measurement of financial risks for regulatory purposes. One objective was to reduce the procyclicality in risk estimates, ie their tendency to be low in tranquil times and spike under stress conditions. Another was to ensure that actual risks, rather than measurement practices, are the main driver of differences in risk estimates across banks. Against this backdrop, we evaluate banks' estimates of credit risk – which sits at the core of bank business models, has featured prominently in banking crises, and seems to have built up during the era of low-for-long interest rates.


To assess the credit risk estimates that underpin banks' compliance with regulatory capital requirements, we use the best available cross-jurisdictional data on large internationally active banks. We juxtapose banks' confidential regulatory credit risk estimates – as collected by national supervisors and compiled by the Basel Committee on Banking Supervision – with vendor data on the same banks' actual credit losses. We also discuss potential inconsistencies between the two data sources and propose remedies. Our data set covers 65 institutions, including 26 global systemically important banks (G-SIBs), from 2009 to 2022.


We find that banks' risk estimates do not accurately forecast the evolution of credit losses but they do rank-order these losses across institutions well, while features of banks and the macro-financial environment help to explain forecast errors. The performance of credit risk estimates along the time dimension reflects their through-the-cycle nature, which is consistent with policy efforts to address procyclicality and helps explain the conservatism in Basel III capital requirements. The accurate rank-ordering across banks is consistent with supervisory initiatives in recent years to assure better measurement practices. The analysis of estimation errors reveals that more profitable banks tend to be more optimistic in their loss forecasts and that banks tend to ignore the credit-to-GDP gap as an indicator of financial overheating.


Focusing on credit risk, we compare banks' expected loss (EL) rates, collected confidentially by the Basel Committee on Banking Supervision from 2009 to 2022, and the corresponding actual loss (AL) rates, as reported in vendor data. Consistent with the use of through-the-cycle risk estimates for regulatory purposes, EL rates rarely move in line with AL rates over time, which helps explain a large precautionary element in Basel III capital requirements. We also find that the rank-order of EL rates across banks matches closely that of the AL rates, in line with recent and forthcoming regulatory efforts to improve risk-measurement practices. EL rates are more likely to be excessively optimistic on the heels of higher bank profitability and financial overheating, as captured by the credit-to-GDP gap.

JEL classification: G21, G28, G32, G33, E44, P52

Keywords: expected loss forecasts; regulatory capital; portfolio credit risk