Global model of banking supervision

Banks must meet a growing volume of complex and granular reporting requirements. The methods used to collect reporting data are not fully harmonized, often leading to redundancies, overlaps, and inconsistencies in the resulting data sets. In this context there is a need to integrate existing statistical data requirements, as far as possible, into a single, standardized reporting framework that would be applicable throughout the euro area and could also be adopted by other European countries, thus easing the burden of bank reports and improving the quality of the data reported to the authorities.

The project scope covers the main supervisory reports: the ECB’s granular credit and credit risk data collection (AnaCredit), the ECB’s stock trend statistics (SHS), the balance sheet items statistics of monetary financial institutions (BSI) of the ECB, as well as statistics of interest rates of the monetary financial institutions of the ECB (MIR); likewise, other needs are considered as a requirement, such as the balance of payments and national accounts.

Our team carries out the project based on persistence in a big data lake, with ingests and transformations of the information in scala / spark technology. The transformation rules to create additional information have been defined based on “Validation and Transformation Language (VTL)” with minor adaptations to improve ease of use and end-user readability.

The generated system that follows this process begins by feeding the input layer from the banks’ internal IT systems, following the structure of the defined input layer cubes. The data processes can be divided into layers (input, enriched layer and output layer) and three phases that separate those layers: transformations in the input layer, output layer and the assignment application that describe the relationships between them. All phases can include validations, for example validations that ensure the integrity of the input layer, validations that ensure the consistency of the enriched input layer, validations that apply the validation rules (external) in the non-reference output layer.