ECB rate risk stress test renews fears over internal models

Banks alarmed by short timeline and opaque supervisory use of IRRBB stress test

The European Central Bank (ECB) has sparked concerns among the banks it supervises that its 2017 stress test, which focuses entirely on interest rate risk in the banking book (IRRBB), could be used to enforce a de facto Pillar 1 capital charge similar in impact – if not in design – to the one rejected at Basel last year. 

Many bankers breathed a sigh of relief when Risk.net first carried the news in January 2016 that the Basel Committee on Banking Supervision had abandoned plans for a Pillar 1 capital charge for IRRBB. The banking industry had argued the risk is highly specific to individual bank business models and market environments, and therefore not suitable for a standardised capital requirement.

But, just over a year later, European banks are starting to fear the idea might be back on the agenda, this time via supervision rather than regulation.

“We do not want Pillar 1 to be reintroduced by the back door, calling it Pillar 2 but keeping many elements of Pillar 1. The reason why Basel actually concluded with the Pillar 2 charge is because you cannot standardise measurement of IRRBB and still make much sense of it,” says Denisa Mularova, senior adviser at the European Banking Federation (EBF) in Brussels.

The ECB announced on February 28 that it was carrying out a supervisory stress test on IRRBB, with banks due to submit data by April 5. The results, both quantitative and qualitative, will then feed into each bank’s supervisory review and evaluation process (Srep) in July, potentially leading to an add-on for Pillar 2 guidance (P2G) supervisory capital if the test reveals risks the ECB considers undercapitalised or qualitative deficiencies in risk management.

The test involves six different interest rate scenarios (see graph): the end-2016 rate curve as a baseline; two 200 basis point shifts in the interest rate curve (one up, one down) that have been used in IRRBB regulation for some time; the end-2010 interest rate curve (more than 200bp above the 2016 curve); and two non-parallel moves, one with the curve steepening, and one with the curve flattening and inverting at the short end.

There are three novelties in these scenarios: the two non-parallel moves, and the fact that the 200bp downshift takes short-term rates well into negative territory at more than –2%. These changes are largely in line with the Basel Committee’s final standards on IRRBB, published in April 2016.

However, they go well beyond existing EU rules, which are embodied in guidelines issued by the European Banking Authority (EBA) in May 2015. Those guidelines specified a zero rate floor for the downside interest rate scenario, and several regulators of nations such as the Netherlands have until now explicitly instructed banks to floor their IRRBB assumptions at zero.

Shock and awe

The ECB indicated last year that it would begin to look at the impact of negative rates on IRRBB, but bankers have been shocked at the speed and manner in which this new approach has been implemented. Legislation that would include the new Basel IRRBB standards – including negative rates and non-parallel shock scenarios – in the EU’s revised capital requirements directive (CRD V) is still winding its way through the drafting process. In any case, Basel does not require adherence until 2018.

Instead, banks are effectively being asked to model risks today based on rules that have not yet been codified in European law, and with a very truncated notice period. The ECB announced the stress test without prior warning on February 28, with a five-week deadline.

“Every bank I talk with is heavily struggling to answer the ECB’s request in a proper way, especially given this very tough timeline. There was no advance notice and this is rather untypical. The question is whether the ECB did this on purpose, to check what is on the shelf at the banks, if they are already fulfilling all the guidelines. After performing this exercise, they will know,” says Tim Breitenstein, a senior manager at consultancy KPMG in Munich.

For many banks, the problems are twofold. First, the stress test coincides with numerous other data-collection exercises by the ECB.

These include quarterly internal capital and liquidity adequacy assessment processes (Icaap and Ilaap), a quarterly short-term exercise whose specifications change each time, regular ‘deep-dive’ workshops carried out by joint supervision teams (JSTs) at individual banks, a profitability test for those deemed domestic systemically important banks (D-Sibs), and a liquidity exercise. The ECB apparently asked for internal audit departments to validate this last item, which is unusual because the second line of defence – risk management and compliance – is the normal location for carrying out this type of supervisory risk measurement. In addition, eurozone banks have faced data requests from the Single Resolution Board.

“If they were given five weeks only on this [IRRBB stress test] and they had to do nothing else, probably the banks would be okay, but it is the same teams doing the ECB reporting, the Icaap, the Ilaap, the liquidity reporting, and this comes in a period when they are actually asked for multiple reporting at the same time by the same teams,” says the EBF’s Mularova.

Splitting headache

The second challenge is that for each scenario, the ECB has requested further slicing and dicing of the results. Banks had to produce separate figures comparing all of their behavioural assumptions for items such as non-maturity deposits (NMDs) and mortgage prepayment risks with the contractual terms; including and excluding macro and micro hedges, and including and excluding a zero rate floor for NMDs in the negative rates scenario.

“In the end, you are calculating for maybe a dozen or more scenarios, and that is definitely a situation banks are moving towards, but quite a number are not there yet, so they will definitely have some issues in providing all the figures that are required in this stress test, especially given the timelines. If they have three months’ time, there is not a problem because they can go to the source systems of all the entities they have and get all the data. But since they need to do it at such short notice, that will be a challenge, and also consistency between the figures will be challenging,” says Erik Vijlbrief, a senior manager at risk management consultancy Zanders in Amsterdam.

The idea behind splitting out these different methods is for the ECB to be able to identify which banks run particular risks. For instance, those especially reliant on behavioural models for keeping IRRBB within acceptable bounds therefore run high model risk, while those especially reliant on derivatives hedging are potentially running extra counterparty risk. The ECB intends to examine counterparty diversification for those banks where derivatives hedging is significant. This adds extra complication, according to the EBF’s Mularova, because counterparty risk data is not usually compiled by IRRBB management units.

Franck Leroy

Franck Leroy, head of financial risks at France’s Groupe BPCE, says once a bank has the data, splitting out different factors to meet the ECB’s requirements is relatively straightforward. “Once you have the data for each product, you can decide to carve out the ones that are linked to models et cetera. So it is just calculation – it takes time, but there is no feasibility issue on this,” he says.

On the other hand, BPCE is a mutual organisation composed of 38 member banks, so the data-collection process itself has proven time-consuming. Moreover, BPCE is one of the larger banks involved in the exercise – it has been designated a global systemically important bank (G-Sib) by the Financial Stability Board. This means regulatory compliance has been fairly front-loaded, and Leroy says BPCE’s convergence with the April 2016 Basel IRRBB standards is already “relatively complete”. This includes the use of a negative rates scenario as part of the bank’s asset and liability management (ALM) modelling process.

By contrast, most of the 110 institutions subjected to the stress test are not G-Sibs and their level of preparedness is likely to have been significantly lower. Vijlbrief at Zanders says the inclusion or exclusion of hedging could be a particular challenge for D-Sibs. Most banks would hedge the interest rate risks on their commercial books internally, facing the bank treasury, which then hedges its residual risks into the market. The hedging positions will need to be reintegrated to identify specific IRRBB hedges and exclude them from the stress test results.

“I think in the end this information is all available, but it is not in the nature of reporting at the moment. Then it comes down to how flexible your systems are. Do you have one central system where you collect all the information? Then it will be quite easy to get this information reported. On the other hand, if you are dependent on local systems where these scenarios need to be implemented, and you are dependent on the set-up of those local systems, a lot of banks will not be as sophisticated in all their entities,” Vijlbrief says.

Moreover, flooring deposit rates at zero in the negative rates scenario could prove highly punitive, especially for net interest income (NII), because banks will be giving away margin on their assets. There is general agreement on the requirement being necessary, however, because several European countries such as Belgium explicitly forbid banks from charging negative deposit rates to customers.

“How does the value of your NMDs change if [rates are] floored at zero? In Europe, maybe 80% of the banks have a model in place where there is no floor implicitly or explicitly included, so they are not able to answer that question. If they put the effort in and answer this question precisely, a lot of banks and the ECB might be afraid of the consequences,” says KPMG’s Breitenstein.

Marginal advantage

Vijlbrief at Zanders says there is one area where the ECB has been relatively lenient – the inclusion or exclusion of margins in economic value of equity (EVE) calculations (see box: Competing metrics). The discounting rate is always the risk-free swaps rate, but many banks are still reporting the cashflow on their assets inclusive of profit margins and credit spreads. The ECB would prefer to see assets listed both including and excluding margins, but has permitted banks to fill in only one field if they can explain the reason. Banks only including margins on cashflow may show more favourable results, but could also become outliers facing additional supervisory scrutiny.

Erik Vijlbrief

“If you read a bit between the lines of the Basel standards provided last year, they state [that] ideally you should, in an EVE metric, create a calculation where at least the cashflow you include is consistent with the discounting you use. So you either use risk-free on both ends or risky on both ends, but mixing it up really results in weird figures. You can see banks would opt for a situation where they would report the EVE with full risky cashflow and then discount at risk-free – they would definitely show quite significant EVE numbers,” says Vijlbrief.

KPMG’s Breitenstein says banks that had not captured the corresponding risk-free rate on their assets at the point of origination would find it very hard to exclude margins for the stress test in a precise way. They would theoretically need to repopulate the data by going back to the point where each contract was signed and identifying the correct swap rate at that time.

“Some ALM systems are already able to do so and have this transparent view, but a lot of systems do not,” he says.

Benchmarking pain

Given such wide variations in methods and capabilities across the 110 banks under scrutiny, there is inevitable anxiety about the ECB’s indication it will challenge bank models and use both quantitative and qualitative results to set P2G capital as part of the Srep. The process of challenge is likely to involve benchmarking results for each scenario output across the sample of banks.

“It is a pretty useful exercise to understand how the bank views itself and how quickly it can gather [data], so on a standalone basis it is quite useful to understand how the bank functions. I think it will be difficult to benchmark from one bank to another,” says BPCE’s Leroy.

Breitenstein suggests three possible ways this benchmarking might take place. First, the ECB could simply compare banks’ inputs with each other and look for any outliers – for instance, how quickly banks assume rates on NMDs will adjust. Secondly, the ECB could compare the sensitivity of outputs to specific modelling assumptions using the data they have asked banks to split out.

“So if you simply compare this model view with a contractual view, where you actually just use contractual data, you will see a big gap. I think this gap is what the ECB will look at, and you can see this as an approximation of the model risk. My guess is if they see there is a lot of model risk at a bank, they may come up with further capital requirements and capital guidance,” says Breitenstein.

On a standalone basis it is quite useful to understand how the bank functions. I think it will be difficult to benchmark from one bank to another

Franck Leroy, Group BPCE

Thirdly, they could compare modelling assumptions with the standardised EVE approach set out in the April 2016 Basel paper. This includes, for example, three NMD buckets with set maturity assumptions: transactional retail deposits (those used for salary payments, for instance) with a five-year average maturity cap; non-transactional retail deposits (4.5 years); and wholesale deposits (four years).

None of this potential benchmarking fills bankers and industry bodies with much confidence. Risk.net has seen a letter, dated March 17, sent to Single Supervisory Mechanism (SSM) head Daniele Nouy by the German Banking Industry Committee of five national banking associations outlining their concerns.

The letter states: “In our view, the possible variation in results owing to differences in assumptions will not provide a suitable basis for drawing conclusions about the quality of banks’ business models (Srep element 1). An assessment of complex internal models must take into account the individual business model and should be performed in a qualitative manner. A supposedly planned benchmarking would not consider the peculiarities of the credit institutions in an adequate manner.”

Even trying to build a harmonised benchmark across all the eurozone markets prompts scepticism. Vijlbrief points out that contract law for items such as mortgages varies country by country, changing the customer behaviour in each market.

“What you see for large European banks is they have quite different exposures in their portfolios in different countries, because of the local regulations or the way contracts are structured. So really comparing across geographies will be quite a challenge, but I think [the ECB] will still attempt it – and that is why they split out the effects from NMDs, from mortgages, and they also split out the hedging, so they get a clear view of what is really the underlying outright interest rate risk that a bank has,” he says.

No shopping list

The controversy over the use of benchmarking matters, because the way the ECB interprets the results will then feed into P2G. BPCE’s Leroy hopes the quality assurance period built into the stress test between April and June will allow a proper dialogue between the JSTs and the banks under their authority prior to the Srep, because there will be specific aspects to the results of each bank that need further explanation to supervisors. These individual results will not be released publicly.

“In France, you have fixed-rate mortgages with quite big prepayment risk, whereas in other countries you do not have the same risks. I hope the ECB will have a better understanding of the business models of the French banks. It is quite different; you don’t have the same risk as others – you have the behavioural risk. However, the amount of credit risk on retail is very low [because of the fixed rates],” says Leroy.

The EBF’s Mularova emphasises there should be no automatic feed-through of stress-test outlier results to P2G. If there were, this would lead the ECB quickly back towards a kind of standardised capital charge, which the Basel Committee rejected when it removed IRRBB from Pillar 1. The EBA has already faced accusations in the past of trying to develop standardised Pillar 2 approaches that are essentially Pillar 1 in all but name. Mularova says the EBA’s 2015 guidelines at least emphasised an outlier reading should only be the trigger for a discussion, not an automatic capital charge.

Denisa Mularova

“Without any room for bilateral discussion, the ECB analysis could lead to some mechanical conclusions; for instance, forcing banks to use these benchmark models without considering the idiosyncrasies of the bank or the type of model. That would be a concern,” she says.

Part of the problem is the absence of any detailed information on how the SSM calculates Pillar 2 capital add-ons in the Srep. Bankers say they believe the ECB has an internal manual on Srep running to hundreds of pages, but little of this has been shared with the industry. The ECB public Srep methodology booklet from 2016 is a fairly general, 44-page PowerPoint presentation in a large font with many diagrams. By contrast, the Bank of England’s (BoE) approach to setting Pillar 2 capital is laid out in a policy statement and a supervisory statement totalling more than 70 densely typed pages between them, providing a detailed methodological breakdown for different risk categories.

One industry source asked Korbinian Ibel, a director-general of microprudential supervision at the ECB who oversees the stress test, whether the SSM would provide banks with a breakdown of Pillar 2 charges. According to the source, Ibel’s response was the ECB would not offer banks a “shopping list”, to avoid the danger of them focusing only on the items that triggered the Pillar 2 add-ons, rather than maintaining a rigorous approach to all aspects of risk management.

They might have been happy a year ago, when they saw it is not Pillar 1, but now it is Pillar 2 and the effects might be even worse for them

Tim Breitenstein, KPMG

The ECB declined to comment on this conversation. However, its explanatory document which accompanied the stress test made clear: “The results of the exercise will feed in a non-mechanical way into the Srep 2017.” Moreover, Risk.net understands that Pillar 2 charges generally are not fixed or automatic; for instance, a repeated governance failure would attract a higher charge than a one-off error.

“The banks have to handle capital charges and requirements for IRRBB. They might have been happy a year ago, when they saw it is not Pillar 1, but now it is Pillar 2 and the effects might be even worse for them. It is simply less transparent, from the banks’ point of view, how the capital charges are derived, but they need to fulfil them anyway,” says KPMG’s Breitenstein.

The fear of a European approach that morphs into standardised Pillar 1 capital requirements for IRRBB has been further spurred by the proposals in CRD V. The directive’s amended Article 84 would delegate to the EBA the mandate to draft regulatory technical standards setting out criteria for the evaluation, identification, management, mitigation, assessment and monitoring of IRRBB by the banks.

“During the Srep, the authority is assessing the internal processes of the banks, which means the EBA is trying to explain to the banks how they should measure the risks in their own internal processes. This makes no sense because this is no longer an internal process – this is a process of standardisation that is absolutely not appropriate. In that case, you can take the ‘I’ out of Icaap and Ilaap because there is no longer any internal process,” says the source.

The EC’s draft would also give the EBA a mandate to set out the criteria for deciding whether a bank’s IRRBB internal models were unsatisfactory, and to propose a standardised methodology that the bank would then be required to adopt instead.

KPMG’s Breitenstein sees the stress test preparing the ground for these new supervisory powers contained in CRD V. “This exercise is an important tool for the ECB to make up their mind whether they trust a bank’s model or not. If they don’t, they will say you have to apply the standardised model. So in this sense, there is a soft push for standardisation,” he says.

Competing metrics

One important change introduced in the Basel Committee standards on interest rate risk in the banking book (IRRBB) in April 2016 was the requirement for banks to calculate shock scenario impacts using both economic value of equity (EVE) and net interest income (NII) metrics. The European Central Bank’s (ECB’s) stress test is effectively forcing banks to front-load this dual measurement. Global systemically important banks (G-Sibs) have generally already shifted toward this approach, but they do not necessarily use both metrics as management tools.

“We are managing our bank more in terms of NII than in terms of EVE. What the senior management of the bank has in mind is the interest margin and how it might evolve over the coming years, so that is a measure everyone knows well in France,” says Franck Leroy, head of financial risks at France’s Groupe BPCE.

He is sceptical about the value of EVE as a management tool, because the outputs depend substantially on what assumptions are made about the duration of non-maturity deposits.

“We [have] calculated EVE for the Basel ratio for a long time now. So it is not a new thing for us, but it is volatile because of residual positions in our balance sheet. This is largely due to the convention you use for things that are not maturing, but I am not sure that is really a risk for us – I’m not linking the volatility of this ratio to real risk,” he says.

The situation in Germany is rather different, according to Tim Breitenstein, a senior manager at KPMG in Munich. Local regulators encouraged banks to adopt EVE during the 2000s, and the rebirth of NII has only really begun with the May 2015 European Banking Authority (EBA) guidelines on IRRBB and the final Basel standards in April 2016. This means domestic systemic banks are likely to be underprepared for the stress test.

“Supervisors said economic value risk is great – you have a full view of your current balance sheet until infinity; you see all your risks. Then the EBA and the Basel Committee said net interest income risk is also very important, it is not enough to look only at economic value. A lot of banks in Germany have problems giving proper NII forecasts and risk numbers,” he says. 

The Basel Committee’s final standardised outlier test for IRRBB is still based purely on the EVE approach. Any bank experiencing a change in EVE of more than 15% of common equity Tier 1 is potentially in line for additional supervisory intervention. However, the rubric of the ECB stress test suggests both EVE and NII outputs will feed into the supervisory review and evaluation process. Amendments to Article 98 in the revised draft of the capital requirements directive (CRD V) proposed by the European Commission in November 2016 also call on the EBA to consider whether an NII outlier test would be appropriate as a threshold for additional Pillar 2 guidance capital requirements.

“What we think is that having two measures for the standard outlier test would increase the complexity of the supervisory process and send out contradictory signals. As we do think EVE will stay, we would be fine with it as the only measure for the standard outlier test,” says Denisa Mularova, senior adviser at the European Banking Federation in Brussels.