The European Central Bank (ECB) has explained how the use of “top-down” as well as “bottom-up” perspectives informed its input to the European Union’s most recent stress tests.
The 2016 stress tests of commercial banks in the EU were run overall by the European Banking Authority (EBA), with the ECB’s supervisory arm helping to examine the banks and to set the parameters used in the process.
The 2016 tests were seen as being “constrained bottom-up” stress tests by many commentators, a chapter in the ECB’s most recent macro-prudential review notes. But the stress tests’ methodology obliged banks “to use certain benchmark parameters and haircuts which were binding for all banks participating in the exercise”, the ECB notes.
The “bulk of these explicit benchmarks were calibrated” using projections calculated via the ECB’s top-down models. The benchmarks were agreed by the EBA board of supervisors after consultation with the ECB and national supervisory agencies.
Combining top-down approaches with bottom-up methods has a number of advantages, the article argues. A top-down approach to quality assurance is model-based, meaning that the same set of models is used to produce benchmark projections for all banks tested. This approach gives banks “a level playing field”.
The fact that the benchmarks are directly linked to the models used to calculate the scenarios used in stress tests means that quality assurance provides a “very good complement” to the judgement of supervisors of particular banks. Top-down quality assurance is also “a data-rich approach”, using information from the entire sample of banks, improving the statistical robustness of the tests.
Banks may face incentives to underestimate the risks they face, the article notes. If that happens, comparisons with other banks “may only be able to identify banks that deviate sufficiently from their peers’ projections without uncovering any bias in the average projection”.
By contrast, the article argues, “the [top-down] modelling approach takes into account model and parameter uncertainty”, allowing supervisors to perform the “evaluation and use of the tails of parameter distributions”. As a result, the estimates formed by top-down models may therefore “be more conservative than the banks’ projections”.
The ECB used top-down models to formulate benchmarks for six types of risk driver: credit risk; net interest income; market risk; operational risk; bank capital projections; and other profit and loss. The main fields where the ECB’s models changed the projections derived from the bottom-up approach were “net interest income and credit risk (relating to loan loss provisions), followed by market risk”, the article notes.
Top-down results from ECB models were “used to challenge bank projections” for credit risk. The article notes that, “given the very prescriptive nature of some of the benchmarks (as published by EBA), these resulted directly in a compliance request in the absence of a credible model provided by the bank in question”.