JP Morgan’s ‘London whale’ losses spark VAR debate

In-house probe will look at role of internal model change, among other factors

The $2 billion trading loss announced by JP Morgan Chase on May 10 has sparked debate about the role played by value-at-risk – with some experts casting VAR as an innocent bystander, while others see it more as a willing accomplice.

Today the loss prompted the bank to announce a reshuffle in which Ina Drew, head of the loss-making chief investment office (CIO) unit and a 30-year veteran of JP Morgan, leaves the firm. The bank is also carrying out its own investigation. A spokesman for JP Morgan declined to detail the scope of that probe, but it is understood that changes made to the VAR model will be one of several factors included.

During last week’s conference call to discuss the loss – an estimated $800 million for the first quarter of this year when offset against the rest of the CIO portfolio – analysts were told the unit changed its VAR model at the start of this year, only to change it back again at the end of the first quarter. When the old model was reinstated, JP Morgan’s average VAR for the first three months of 2012 leapt to $129 million – almost double the $67 million disclosed in the bank’s first-quarter earnings. Average VAR in the fourth quarter of 2011 had been $69 million, implying either a big change in market conditions over the first three months of this year, or a big change in the position itself.

That has raised eyebrows – and suspicions. “The cynical view is that the guys putting on trades figure out the weaknesses in the new VAR model, and put on positions that do not result in increased modelled risks,” says Christopher Finger, executive director at MSCI in Geneva, a provider of investment support tools. “You can see under the old VAR model the risk of the position doubled. The market has moved, but I think it’s more likely mostly due to a position change in the portfolio. You have to wonder: if somebody had been looking at the old model’s numbers, would they have taken different actions?”

Others agree changes to the VAR model might have helped obscure the amount of risk taken by the CIO – but claim it could have been an innocent mistake. The model might have been altered to remove a risk factor the bank had decided was immaterial for the business, or that no longer existed, for example. The losing positions have been described by JP Morgan as part of a synthetic credit portfolio – widely reported as a curve strategy that involved buying and selling protection on Markit’s CDX North American Investment Grade index of credit default swaps.

You have to wonder: if somebody had been looking at the old model's numbers, would they have taken different actions?

“They probably thought the CDX trades had too many VAR breaches in 2011, which caused the bank to raise the VAR limit. The new VAR model was data-mined to produce the desired number of breaches in the past, and as a result halved the VAR relative to the old model – this is easy to do. In retrospect, it looks like evasion, but it is possible it was a sincere effort to improve the VAR models,” says the chief risk officer at one hedge fund.

During that process, something must have been missed – possibly resulting in the VAR numbers misrepresenting the relationship between two offsetting positions, one quantitative analyst speculates. “It looks as though they have missed out some risk factor. The risk model contained wrong assumptions on the dependence of what they hoped were mutually hedging portfolios, setting some basis spread to zero, for instance,” he says.

The hedge fund’s chief risk officer illustrates that scenario by pointing to the basis between US Treasury yields and Libor. “Say the old VAR model used the volatility of short-term US Treasury rates as an input. In 2008–09, the volatility of Libor was much higher than the volatility of US Treasury bill yields. So you switch your model to use Libor volatility – which is currently not that much higher than US Treasury bill yield volatility – so it looks as though VAR has declined significantly,” he says.

There is also a third point of view – that VAR is a red herring. Even if model-set limits had not been breached, the losses would still have occurred and the bank should have been using a battery of other metrics and controls, risk managers say.

“VAR isn’t the only thing you look at, especially on this kind of large trade. They should have had concentration and liquidity risk metrics that would have told them the position was risky. It can be complicated to model seriously but even a basic market-share metric would have been a signal,” says a risk and capital manager at one European bank.

A senior risk manager at a US bank puts it more bluntly: “This isn’t because of a modelling problem – don’t be naïve. Whatever the risk measure was over one day would not have made much difference once they had that kind of position. This was a fundamental failure of high-level risk management – they got too big for that market, everybody knew they were, and they got killed,” he says.

On May 3, the Basel Committee on Banking Supervision proposed replacing VAR with expected shortfall as the standard risk metric for trading book capital requirements. But the use of an alternative risk measure would not have made much difference in the JP Morgan episode, experts say.

“Expected shortfall probably wouldn’t have helped here either,” says the European bank risk manager. “It would be a bit more conservative because it looks a bit further into the tail, but that only really makes a large difference for discontinuous risk exposures, which doesn’t seem to be the case here.”

In the reshuffle announced today, Drew will be replaced as CIO head by Matt Zames, currently co-head of global fixed income in the firm’s investment bank and head of capital markets for JP Morgan’s mortgage bank. Zames will also join the bank’s firm-wide operating committee. His global fixed income co-head, Daniel Pinto, will now be sole head of that group. Pinto remains chief executive for JP Morgan’s businesses in Europe, the Middle East and Africa.

The bank’s response to the losses will be led by Mike Cavanagh, chief executive of its treasury and securities services division.