Banking supervision requires regular inspection and assessment of financial institutions. In Germany this task is carried out by the central bank ('Deutsche Bundesbank, BBK') in cooperation with the Federal Financial Supervisory Authority ('Bundesanstalt für Finanzdienstleistungsaufsicht, BaFin'). In accordance with the Basel II approach, quantitative and qualitative information is used. It is still an open question whether supervisors provide information, based on on-site inspections, which is not known from the numbers already, or simply duplicate the quantitative information, or even overrule it by their impressions gained through visits. In our analysis we use a unique dataset on financial institutions' risk profiles, i.e. the banking supervisors' risk assessment. Methodologically, we apply a partial proportional odds model to explain the supervisor's ordinal grading by a purely quantitative CAMEL covariate vector, which is standard in many bank rating models, and we also include the bank inspector's qualitative risk assessment into the model. We find that not only the quantitative CAMEL vector is clearly important for the final supervisory risk assessment; it is, indeed, also qualitative information on a bank's internal governance, ICAAP, interest rate risk, and other qualitative risk components that plays an equally important role. Moreover, we find evidence that supervisors have become more conservative in their final judgement at the beginning of the financial crisis, i.e. the supervisory assessment seems to be more forward-looking than the mere numbers. This result underpins the importance of bank-individual on-site risk assessments.
