How banks game stress tests: the ‘shocking’ truth

Leaked memo exposes effort to swap out risky assets despite Fed’s push to end “window dressing”

Risk 1019 Lead story Stephen Lee NB illustration
School of (not so) hard shocks: banks stand accused of manipulating the market shock scenario
Stephen Lee, nbillustration.co.uk

  • The Federal Reserve’s global market shock scenario has a floating start date, to deter banks from gaming the test by temporarily swapping market risk off their balance sheets.
  • However, Risk.net has seen evidence at least one US systemic bank is still looking to “window-dress” its balance sheet around the stress test.
  • The European Banking Authority has proposed a floating start date for its 2020 stress test, bringing it in line with the Fed.
  • But European bankers say the EBA test is not as binding as the Fed’s, so there is less incentive for window dressing.
  • Experts believe existing regulatory and risk reporting data could be used to identify and prevent window dressing.

It seemed innocuous enough. A large US bank approaches a European dealer proposing to swap equities for bonds on a triparty basis, in an arrangement it describes as an “upgrade trade”.

Later discussions, detailed in a memo dated May 2018, show the underlying purpose of the trade: to help the US bank with “liquidity needs during stress test periods”. In other words, the bank wanted to massage its trading book in the lead-up to the US Federal Reserve’s annual stress test by temporarily offloading market risk.

The memo, leaked to Risk.net, shows the practice of balance sheet “window dressing” is alive and well despite regulators’ best efforts to stamp it out.

Financial authorities conduct stress tests to gauge banks’ resilience to market shock and determine the amount of capital they must hold. If a dealer can game the tests to show an artificially low picture of risk-taking, it may end up with insufficient capital to withstand the next crisis. The risk is magnified if the bank is a large, systemically important one.

“The concern would be: are banks taking actions specifically to manage against the stress-testing regime that actually increase their risks or increase risk to the system?” says a former Fed supervisor. “It depends on the size as to whether this is a systemic concern. It is clearly a potential systemic concern, in the sense that it may mask the true risk positions of a systemically important bank.”

The US bank in the memo is one of eight US G-Sibs, or global systemically important banks. They are among the 11 banks subject to the global market shock scenario, part of the Federal Reserve’s annual stress test, known as CCAR.

The scenario is designed to capture risks in the trading book. To deter banks from window-dressing their books, the Fed applies a moving reference date for the test. Up to 2016, the range for that start date was any time between January 1 and March 1 of the year in which CCAR was taking place. However, the Fed didn’t want to use a date too close to the early April submission deadline for capital plans, because banks might not have time to prepare and check the data. As a result, the so-called “as of” date tended to be early in January.

And that predictability created a problem. In a September 2016 consultation on proposed reforms to CCAR, the Fed observed: “The narrow window creates the possibility for bank holding companies to artificially reduce the risk of their portfolios around the time of the market shock date.”

To restrict this arbitrage, the regulator proposed widening the range of possible “as of” dates. In the revised rule finalised in February 2017 and applied from the 2018 CCAR onwards, the Fed adopted a range starting on October 1 instead of January 1. In the 2018 test, the date used was December 4, 2017, and for this year’s test, it was November 5, 2018.

The contents of the memo indicate the wider window has not completely deterred banks from trying to window-dress their trading books around the timing of the market shock scenario.

That’s a worry. The Fed is being pressured to increase the transparency of CCAR models, in order to help banks improve their capital planning as part of the rollout of the Fed’s proposed stress capital buffer – which will be based on a bank’s CCAR results. Former Fed governor Daniel Tarullo has already suggested, during a seminar in Washington DC on May 21, that more transparency allows more opportunities for large banks to finagle the tests.

“If the stress test becomes predictable, it ceases to have value. I suspect the smart people who work on these things have most of what they need to reverse-engineer the models,” Tarullo warned.

European banks are playing their part in the practice, perhaps encouraged by the bloc’s laxer stress-testing regime. The memo talks of “mutual benefit” in the proposed asset swap trade, hinting the US bank could return the favour and help the European bank to window-dress its own trading book, presumably in the lead-up to next year’s stress test conducted by the European Banking Authority.

The EBA’s head of stress-testing, Angel Monzon, admits there is a need to make sure banks cannot game the tests. The authority is now planning a clampdown of its own that could be even tougher than the Fed’s approach – but only once every two years.

But critics say a floating start date is a primitive way to tackle window dressing. Even the bankers themselves think regulators could clamp down more effectively, especially by making better use of the vast quantities of prudential reporting data they already receive.

Anatomy of a trade

For the “upgrade trade”, the US bank offered to post equities on a triparty basis, and receive bonds under a bilateral securities lending master agreement. This kind of asset swap activity is entirely permissible within the rules of the stress test, and the trade was discussed by the European bank’s prime financing division with a view to identifying potential hedge fund clients to take the other side.

The timing of the transaction suggests the US bank was seeking an upgrade trade in the fourth quarter of 2018, to help buff up its 2019 CCAR results. A source at one bank says their repo desk noticed a premium on equity swap trades during the fourth quarter of 2018. That could indicate either banks actively seeking to swap equities off balance sheet, or (more innocently) an unwillingness among banks to take on extra equity exposure until the market shock start date was known.

One senior stress-testing manager at a global bank says his firm would be reluctant to engage in window dressing for fear of a hit to its reputation. But he acknowledges there is an incentive to do so in the US, given that CCAR is – for many large dealers there – a binding constraint that directly affects capital management.

Marcus Stanley, policy director at pressure group Americans for Financial Reform, points to circumstantial evidence that banks have been tightly managing the effects of the CCAR market shock.

Trading losses and counterparty losses under the severely adverse scenario have been strikingly consistent, running at between 110 and 120 basis points of Common Equity Tier 1 in aggregate, every year since 2014, with the exception of a lower loss in 2017 (see figure 1). Stanley notes this impact is surprisingly low, given the actual size of capital depletion that resulted from trading and counterparty losses during the financial crisis in 2007 and 2008.

“They are always telling us how extreme the shock is – are you really saying all the banks would lose just $100 billion on an extreme market shock? So there are reasons to be suspicious of the market shock,” Stanley says.

Asked for comment, the Fed referred Risk.net to the September 2016 consultation document and the February 2017 revised final rule.

Win-win (except for supervisors)

Window dressing is not limited to stress tests. Banks dabble in the practice to improve their end-year and end-quarter leverage ratio, especially by shrinking repo books; or employ collateral upgrade trades to help improve their quarterly liquidity coverage ratio.

But that activity revolves around a fixed reporting date, which means banks would know in advance what start and end dates to set for the trade. The introduction of a much wider range of possible start dates since the 2018 CCAR means banks would have to enter the trade at the start of October, with the timing of the exit anywhere until early the following year.

One equity structurer in Europe says the unknown end date for the trade would complicate matters. The pricing and hedging is already complex, due to the different costs of carrying the two instruments being swapped, in terms of relative liquidity and cashflow risks from bond coupons versus equity dividends.

For many bankers, though, any price premium on the trade would be worth paying if it helped the firm ace its CCAR test.

“The more you are constrained by CCAR as a bank, the more you want to get rid of the equity exposure, the less concerned you are about pricing the trade accurately,” says the equity structurer.

He says he has not been presented with anything expressed as frankly as the memo seen by Risk.net. Most banks would shy away from marketing trades as linked explicitly to CCAR, he says, and try to offload the risk in a more clandestine manner.

“Most likely they would just break up the trade and sell it to second or third tier banks in Europe that are less aware of CCAR and the reputational risk of being seen to go against the spirit of regulation. [The G-Sib] would just pitch it as a good yield pick-up on the counterparty’s bond portfolio,” he says.

Because banks know about the stress test and they know they need to apply the scenario with a cutoff date as of the end of the year, they may change the composition of their market risk portfolio

Angel Monzon, EBA

The design of CCAR also makes it likely that banks would break up window-dressing trades into smaller pieces, the former Fed supervisor says. For the largest and most interconnected dealers, the adverse and severely adverse components of CCAR include a counterparty default scenario, with the same “as of” date as the market shock scenario.

If one of those dealers undertakes a large window-dressing asset swap with a single counterparty, it could end up being captured under the CCAR counterparty default module, undermining the whole purpose of the trade.

The equity structurer adds that hedge funds are also keen recipients of exotic correlation risks that banks want to offload, as potential new sources of alpha in a market where trades quickly become crowded. A second former Fed staffer says the regulator was aware of banks purchasing put options for extreme events from non-banks, which could be counted as a hedge when running the global market shock scenario.

Whichever method is used, the pricing could work for both counterparties. The hedge fund or receiving bank would work out a price for the trade based on current market volatility, whereas the US G-Sib will have one eye on the much higher level of volatility and potential loss implied by the CCAR global market shock scenario. If the stressed capital saving is sufficient, says the structurer, it would justify the G-Sib paying over the odds to swap its equity exposure off balance sheet.

The European connection

The “upgrade trade” is one leg of a possible quid pro quo. A reciprocal trade would see the European bank benefit from an asset swap in advance of its own stress test. Compared with the US, this would be a relatively easy task. In previous tests, the entire shock scenario run by the EBA has started on December 31, including the market risk component. Hence EU banks could rely on a very short-duration trade with a fixed start and end date.

Like the Fed, however, European regulators are alert to this risk. In a draft methodology for the 2020 stress test, released on June 25 this year, the EBA asked for feedback on the idea of a floating start date for the market risk scenario, to be set any time between September 1 and December 31, 2019.

“Because banks know about the stress test and they know they need to apply the scenario with a cutoff date as of the end of the year, they may change the composition of their market risk portfolio,” says Angel Monzon, the EBA’s stress-testing chief.

He adds: “At the end of the year, banks may reduce their market activity, maybe because of the holiday season and also because of the end-of-year balance sheet presentation of their results. This might impact the representativeness of the market risk portfolio. For all these reasons, we thought about this proposal.”

The actual start date would only be announced when the stress test exercise is launched, typically at the end of January or beginning of February (2020, in this case). That’s tougher than the Fed’s approach, which is to notify banks of the “as of” date no more than two weeks after it has passed. The EBA proposal would therefore mean any window-dressing trade would have to last for a full four months. The equity structurer says this could well make the trade uneconomic.

The EBA’s Monzon emphasises that the methodology is only a proposal at this stage. The regulator is likely to publish its final 2020 stress test methodology in November this year. The EBA applies its market shock scenario to a sample of 50 banks.

The switch to a floating start date could deter not only window dressing by European banks, but also their involvement in trades to help US banks soften the impact of CCAR. If EU banks are no longer sure on which day during the fourth quarter their own trading books will be stress-tested, they will be reluctant to take on extra market risk from US banks during the same period.

However, the EBA only conducts its test once every two years. In the intervening years, the European Central Bank tests the banks under its direct supervision, but this is usually focused on one narrow aspect of banks’ risk – for instance, interest rate risk on the banking book in 2017, and liquidity risk in 2019.

Consequently, in the years the EBA exercise does not take place, European banks would potentially face no stress test implications of their own if they temporarily receive equity risk from US banks. That points to the need for additional co-operation between the Fed and European regulators to try to close the biennial loophole.

“As a former supervisor, I would always argue there needs to be more information-sharing across major supervisors,” says the former Fed supervisor. “By shifting the positions into a European bank, are you creating outsized positions where they could not be well absorbed?”

US banks have a greater incentive to game the market shock scenario than their European counterparts, since the results of CCAR feed directly into their capital requirements. In Europe, the EBA stress test is only used to guide supervisors and help them decide the size of the Pillar 2 capital add-ons for banks. The results do not feed directly into the size of their capital stacks.

This leaves European bankers questioning whether a floating start date for market risk is even necessary.

“Would we even bother [with window dressing]? No, because it would be an impediment to the business, and there are no risk-weighted assets involved, so we don’t save anything,” says the head of market risk at one European bank.

The senior stress-testing manager at a global bank agrees that neither European capital managers, nor investment analysts who follow the banks, pay much attention to the EBA test. He says the impact of the EBA stress test results in terms of investor reaction is “so low, it is crazy to think any trades would be done for window dressing”. He adds: “The effects of the test are very far from the attention of the front office.”

A VAR better idea

A floating start date is not the only tool available to supervisors to curb arbitrage of the market risk scenario. Observers say a more efficient – and more reliable – method involves using existing bank data to check there are no unexplained swings in the size of market risk positions ahead of known or expected stress test dates. Banks already produce this data for their own risk management and for supervisory reporting.

“You could have a scaling factor looking back on value-at-risk in the fourth quarter, or just to check that you don’t have any sharp changes in your risk during the previous three or four month period, to avoid any window dressing. I expect the banks will respond [to the draft methodology] with suggestions like this,” says Hannan Mohammad, a director in financial risk management advisory at consultancy KPMG and former derivatives trader.

The stress-testing manager suggests using European banks’ own internal stress tests, which may cover fewer risk factors than the EBA test but in many cases are run daily. These internal tests are subject to auditing, and are challenged by European Central Bank on-site inspection teams, so “banks cannot just do what they want”, he says.

As a benchmark for its test, the EBA could use the date at which the internal stress tests of the bank produce the largest outcome. Or, it could use an average of a period of, say, 20 days where the results were the largest, says the manager. The EBA would then calibrate the results with a scaling factor derived from end-of-year accounts.

Interestingly, in this instance the views of market risk experts at the banks coincide with those of ex-regulators and the advocates of strong regulation.

The AFR’s Stanley points to a research paper published by three staff at the Fed in November 2018, which used daily trading book profit and loss data to infer market risk factor exposures at 13 banks the Fed supervises. Stanley says this kind of daily data should help the regulator spot if banks are altering the content of their trading books around the start date of the global market shock scenario.

He also wants the Fed to make better use of banks’ reporting under the Volcker rule. If the rule is being implemented properly, Stanley says, this data is supposed to be detailed enough to identify if dealers are indulging in banned proprietary trading, and should therefore be usable for other supervisory purposes as well.

“The Volcker rule is a reporting regime at the trading desk level. Banks and supervisors are supposed to understand what is happening at – and be getting a steady stream of daily data from – the trading desks. So these kind of [window-dressing] moves should be findable and detectable,” he says.

The former Fed supervisor says evidence of continued CCAR window dressing is a timely reminder of why the stress test needs to be a flexible tool, rather than going down the path of rigid rules-based requirements like the Basel standards.

“It has always been a struggle to keep the stress test dynamic and not to introduce ways in which banks can try to get around it,” says the former supervisor.

Editing by Alex Krohn

Risk.net will be publishing a full interview with the EBA on the future of European stress-testing in the coming days.

  • LinkedIn  
  • Save this article
  • Print this page  

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact info@risk.net or view our subscription options here: http://subscriptions.risk.net/subscribe

You are currently unable to copy this content. Please contact info@risk.net to find out more.

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here: