Q&A: Finansinspektionen's Uldis Cerps on capital floors and too-big-to-fail

At first glance, Swedish bank supervision in the post-crisis years might appear contradictory – the country's central bank and its prudential supervisor, Finansinspektionen, have sought to control mortgage risk through a variety of mechanical means, sidelining banks' risk-weighted asset (RWA) calculations. But the supervisor has also consistently spoken up for risk-sensitive capital requirements, and Uldis Cerps, Stockholm-based executive director of banking supervision, does so again in this interview.

How to make sense of it? Cerps says the mortgage crackdown, which encompasses a loan-to-value (LTV) cap, a 15% risk-weight floor, a subsequent hike to that floor, and now a consultation on amortisation requirements, was a result of the two decades of benign data that followed Sweden's home-loan crisis in the early 1990s. Banks were correctly applying the regulatory parameters, he says, but given the dataset, those models would inevitably spit out low default probabilities. So supervisors had to step in.

Elsewhere, models are more suited to the task at hand, and Cerps argues it is important that a planned system of risk-weight floors – to be set as a percentage of the relevant standardised capital number – does not overstate capital requirements and inadvertently encourage banks to take more risk (Risk February 2015). As a member of the Basel Committee on Banking Supervision, which is running the consultation, Cerps will be able to make his voice heard.

Cerps also discusses ongoing efforts to modernise banks' trading book capital requirements – conceding more thinking needs to be done on plans for models to be approved at the desk level – and efforts to end the too-big-to-fail (TBTF) problem. Banks still enjoy a TBTF premium, he believes, implying the market still thinks there is a chance of a taxpayer bailout; and he also raises concerns about plans for banks to hold enough liabilities to enable recapitalisation – proposed internationally under the heading of total loss absorbency capacity (TLAC) and in Europe as the minimum requirement for own funds and eligible liabilities (MREL). If a breach of the TLAC total is treated as seriously as a breach of minimum capital requirements, Cerps warns a bank could be tipped into resolution simply as a result of a refinancing squeeze.

What have been your main priorities since taking on the role of executive director of banking supervision in May 2008?

Uldis Cerps: In 2008, the priorities were very much linked to de-risking banks in the Baltics, and to ensuring major banks raised more capital in 2009 and in 2010 in response to the crisis. After this, we extensively engaged in the regulation of banks' liquidity. We adopted the liquidity coverage ratio requirements in 2013.

I think banks still enjoy a too-big-to-fail premium, and the market has not yet fully accepted that everybody who may need to be bailed in will be bailed in

We have also pursued more structured policies over the years in response to risks in the Swedish housing market. In 2010, we introduced a cap on mortgage LTV at 85%. Then, in 2013, we introduced the risk-weight floor for banks using the IRB approach at 15%, raising it a year later to 25% for macro-prudential reasons.

We have also decided to maintain the Basel I floor in Sweden and not give waivers to any banks. In addition, on March 11 we launched a consultation on mandatory amortisation requirements to deal with the situation in the housing market from a macro-prudential perspective. We have also addressed in a more systematic way the systemic importance of the four major Swedish banking groups, and we have done that by requiring them to hold an additional 3% capital in Pillar I and 2% capital in Pillar II. That is an additional, and big, measure to address systemic risk in Sweden.

Floors are set to become a standardised part of bank supervision. How might the level of the floors be defined?

UC: The current Basel Committee consultation document, which is still out for public comment, envisages three ways of dealing with floors. There is a choice between just one floor, like today, for all risk types, which allows you to offset one risk type against another. Then you have one floor for each risk type – operational, market or credit – and then the option of applying floors for each exposure class, such as lending to small and medium-sized enterprises or sovereigns.

The Swedish supervisory authority, the FSA, prefers a very granular floor, especially if it's going to serve as a backstop to risk-weighted asset models.

What would an ideal level of calibration be, in order to serve as a backstop?

UC: For us, the most important thing is that if you think the banks measure risk correctly – although they have not always done – then calibrating the floor much above the true risk would create incentives for banks to take more risk. That is not the overall intent of the framework, so I think the calibration needs to be very careful.

Bank risk managers have claimed a floor anywhere north of 75% would destroy incentives to model. Do you agree?

UC: I don't think it's appropriate to talk about percentages if we do not know whether we're talking about the overall floor, risk-class floor, or exposure-class floor. These are three very different animals.

In Sweden, when we wanted to fix the problems with internal ratings-based (IRB) models with respect to Swedish residential mortgage risk weights, we did not think banks had incorrectly applied the IRB models. The problem was we had more than 20 years of benign housing market data, which did not show up any major losses in the residential segment, so we thought the models actually underestimated risk. We subsequently introduced a risk-weight floor equal to 15% for all banks using the IRB model in Sweden. That 15% was the result of very careful calibration and international comparison, and we wanted to make sure we did not incentivise just that type of risky behaviour.

Why was the floor increased to 25% last year, and how did banks respond?

UC: The reason we introduced it in the first place at 15% was to make sure the IRB models generate sufficient capital to address the risks to which the banks were exposed. We raised it from 15% to 25% to ensure the risks banks cause to others are properly reflected, in the sense that the macro-prudential risks are efficiently taken on board.

It's interesting, because when we raised risk weights to 15%, we actually got the major support from almost all industry representatives, I think they realised this was the sensible thing to do. I would not say the macro-prudential increase was welcomed by the industry in the same enthusiastic manner. I think there were concerns we were overshooting, and there were concerns the risk weights are too high in relation to the risk and that we have gone too far.

At the same time, we can see the increase to 25% has not reduced the growth of the lending rate. It was still higher last year than the year before. So, we feel the increase from 15% to 25% was absolutely something that was necessary for us to take, and we have been very clear that there might be a need for additional macro-prudential measures in the Swedish housing. This was confirmed when we consulted on the mandatory amortisation requirements.

Will you take action on the mandatory amortisation requirements this year?

UC: We expect that to come into force via our board decision later this year. The consultation period began last week, and it has been proposed that if banks have customers who take new loans with an LTV ratio exceeding 75%, such loans shall be amortised at an annual rate of 2%, and new loans with an LTV between 75% and 50% shall be amortised at a rate of 1% a year. When an LTV of 50% is reached, it is essentially up to the parties to agree on future terms of amortisation.

The Fundamental review of the trading book (FRTB) is set to be completed fairly soon. Will it meet the year-end deadline?

UC: I have no particular Swedish insight into the timeline, I think that will depend on the overall work of the Basel Committee.

If it is wrapped up by the end of the year, will that have been too quick? Has there been time to fully assess the impacts?

UC: My overall personal view is that it is important to always present a credible response to the crisis, because it has cost the taxpayers a lot of money and has resulted in the fundamental overhaul of both banking and supervisory models. At the same time, it's important that the responses are also sufficiently long term and do not need to be changed too frequently. I think it is a prime balancing act.

How will it be implemented in Europe? Will there be a fifth Capital Requirements Directive?

UC: The traditional way of implementing banking legislation in Europe has been through changes to the capital adequacy framework. Whether it will be a directive or a regulation depends on the choice of politicians and their appetite for maximum harmonisation.

We have heard the Basel Committee will issue new proposals on credit valuation adjustment (CVA) in the coming weeks – which follows work done by the European Banking Authority (EBA) to assess Europe's CVA exemption. Is that right?

UC: I think as a starting point, in line with the view of the EBA, banks that have significant CVA risks must be charged capital, and if for some legal reason this risk is not sufficiently covered in Pillar I, then it has to be covered in Pillar II. If not, it would be contrary to the overall principle that all risks have to be covered by capital in banks.

This was one of the areas where the EU implementation was not considered to be consistent with existing Basel capital standards. The Basel Committee is still working on changes to the existing framework, but that does not take away the obligation of the individual members to actually implement the existing standard as envisaged.

Will we see a new document in April?

UC: I think it's better to check with the secretariat on the deliverables from the specific workstreams.

Returning to the FRTB, what are your views on regulators seeking to replace global approval of value-at-risk models with desk-level approvals?

UC: More thinking needs to be done, because there are different ideas of what you want to attain with the models. One is, of course, to improve risk sensitivity, but the models also need to be sufficiently robust and to generate satisfactory capital outcomes. And if there are too many models in one bank, of course the supervisory resources required to review them may exponentially increase. That is one consideration.

In terms of the threshold for approvals, how reliable will models have to be?

UC: In my view it is premature to answer that question at this point.

Do you have a personal view on where it should be set?

UC: Our thinking in terms of the role of standardised models in the market risk space is lagging a little way behind standardised approaches in the credit risk area, where we have quite firm views. More thinking needs to be done.

How will desks be defined? How many would you expect a big bank to have?

UC: Ultimately, I think that relates to the question of how many models you want to have for a bank, and I think it segues into a much bigger question of what role you see for models in the overall capital framework. We need to make sure risks that are able to be modelled robustly can continue to be modelled, and risks that are not are measured by standardised approaches.

How close would you say regulators are to ending too-big-to-fail?

UC: This is a big issue. Europe's Bank Recovery and Resolution Directive was supposed to be implemented by all countries at the start of this year, and there are a couple of important things to be kept in mind. First, I think banks still enjoy a too-big-to-fail premium, and the market has not yet fully accepted that everybody who may need to be bailed in will be bailed in. But the second observation is that banks are getting much better prepared for avoiding resolution, in the sense that there have been tremendous efforts going into recovery planning, both on the banking and regulatory sides.

In terms of how far we are in practice from implementing the resolution plans, we all want to have a robust resolution regime, so it's important banks have sufficient MREL and that the authorities have sufficient capacity to actually bail them in. At the same time, it's very important to have sufficient buffers on top of minimum requirements to minimise the probability of resolution. Banks in Sweden are required to have a 5% systemic risk buffer, and they will also have the countercyclical buffer of 1%.

In addition, they have Pillar II requirements on top of Pillar I, and they will have to have a capital conservation buffer and be subject to stress testing. In Sweden, our requirements range somewhere from 15% to 19% for banks' Tier I common equity ratio. This will give banks sufficient self-defence. Of course, sooner or later the TLAC requirements will arrive, so I would say the resolution agenda is still a work in progress.

Will Swedish banks need to issue more debt to meet the TLAC standards?

UC: It remains to be seen. First of all, we have now assigned the authority in Sweden to set the MREL requirements, which is the Swedish national debt office. It will be up to them to set the MREL requirements, and they will have to do that this year. TLAC is still in the process of consultation, so we will have to see what the impact assessment shows and how the final calibration looks.

There hasn't been any kind of indication of where it will be set in Sweden?

UC: The debt office got the news that it will be the Swedish resolution authority on March 11, so it's not particularly surprising it has not yet decided on this.

Are there any concerns the demand will not be there to satisfy supply across the industry? Will there be enough buyers?

UC: We are generally very supportive of the idea behind MREL; it offers the possibility to have cushions in times of need for bail-in, and it also potentially minimises the too-big-to-fail premium. From that perspective, it is good. At the same time, MREL may have the same risk we observed with the TLAC proposal – if TLAC debt matures, then you automatically use other types of capital to meet the requirement, and then you can have a hypothetical situation where a bank with quite solid capital adequacy still doesn't meet full TLAC requirements.

In that sense, I think it's important that not only the absolute amounts of MREL are carefully considered, but that also the consequences of not meeting MREL are understood and well thought through – in my view, you do not want banks to fail in cases where they still fully meet capital requirements but fail to meet MREL. You don't necessarily want MREL itself to be the factor that facilitates a more frequent need for resolution. That's one of the points we and the Swedish national debt office have particularly pointed out in our response to the Financial Stability Board TLAC consultation, since we are not represented on the board.