Behind JPMorgan’s Big Loss: A Broken Risk Model

Losing $2 billion hurts. But JPMorgan Chase may ultimately suffer even greater pain due to the fact that it was blind to its exposure long after outsiders caught on. As its Chief Executive, James Dimon, awkwardly noted last week, “stuff in the newspaper and a bunch of other stuff” should have provided the company with ample notice of a problem.

Like many banks, JPM used so-called value at risk modeling to gauge its exposure and warn of threats ahead. The problem is that its VaR model didn’t provide any early warning because they were fundamentally flawed, the company’s first quarter regulatory filing indicates.

Beyond the ensuing embarrassment, questions about JPMorgan’s VaR models and the bank’s revised, elevated readings could pose a threat to its future ability to take on risk in its trading book.

VaR is supposed to calculate, with a 95% chance of being right, the amount a bank’s positions could gain or lose in value during a single trading day. JPM’s first-quarter earnings announcement initially pegged the amount its London-based chief investment office could lose in a day at $67 million, based on a new VaR model it had begun using that same quarter. Shortly after the end of the quarter, however, the CIO’s positions began racking up losses well in excess of that.

When reality didn’t match the bank’s formulas, JPMorgan readopted its old model – which promptly tripled the end-of-quarter VaR to $186 million, according to its first quarter SEC filing.

Suggesting that a flawed VaR formula directly caused JPMorgan’s losses would be unfair: No bank chief risk officer would make decisions based solely on a VaR’s results, given how inadequate they’ve proven in the past.

“It’s not a ‘coherent’ risk measure,” notes Viral Acharya, a professor at New York University’s Stern School of Business. It gives no sense of worst-case scenarios and assumes liquid markets that will allow banks to speedily exit their position without increasing losses.

When a player is large, of course, the risk increases that anything it does will move the market against it. That’s exactly what appears to have happened at JPMorgan, Acharya says.

Unfortunately, overconfidence appears to be a permanent component of VaR modeling – perhaps because it’s often used to placate regulators. JPMorgan’s chief investment office is sitting on a $360 billion portfolio. That means in boasting with a 95% confidence level that the office would not lose more than $67 million in a day, it was claiming the red ink would represent no more than a seemingly quixotic 0.02%. The adjusted level of $186 million doesn’t seem all that more plausible.

Jamie Dimon, who earned renown during the crisis for having trusted common sense over modeling, is no doubt well acquainted with VaR’s limits. But if it wasn’t VaR Dimon was relying on, it’d be good to know what he was using when he dismissed as a “tempest in a teapot” outsiders’ concerns about his chief investment office’s positions.

That question will no doubt interest regulators, which could in turn hamper JPMorgan’s ability to take on risk. As of March 31, the VaR for the chief investment office’s positions alone exceeded $186 million. That’s more than double the VaR reported by the entire company a year earlier and a number that, presumably, roughly matched what the bank and its regulators regarded as a safe limit at the time.

If regulators are relying on VaR, they could be forgiven for asking why JPMorgan should be taking on any risk until it can clear out of the CIO positions. If Dimon has a convincing argument that the bank truly has a better read on its own trading risk than VaR, now would be a great time to trot it out.

Jeff Horwitz is a co-editor for risk management at American Banker. The views expressed are his own.