Value at Risk (VaR) and J.P. Morgan have long been intertwined. The bank pioneered this risk model that purports to show how much money a bank stands to lose on any given day. Proponents argue VaR is a way to effectively calculate risk and spot losses before they happen. The $2 billion trading loss at J.P. Morgan -- with estimated, though unconfirmed, figures reaching as high as $3 billion --which stemmed from a portfolio hedging strategy that relied on credit derivatives but went awry, calls into question many things: CEO Jamie Dimon’s leadership, the effectiveness of the Volker Rule and other financial reforms, to say nothing of the judgment of the actual traders and risk managers directly involved. But it also raises questions about the effectiveness of current VaR models and their future at financial institutions — or so say some analysts.
Dimon himself mentioned VaR models in his conference call late last week about what went wrong at J.P. Morgan. “In the first quarter we implemented a new VaR model, which we now deemed inadequate and went back to the old one that we used for the past several years, which we deemed to be more adequate,” said Dimon, cryptically. He added, “There are constant changes and updates to models; we’re always trying to get them better than they were before. It’s an ongoing procedure.”
One interpretation of what Dimon may have been getting at regarding VaR, and the specific weakness in the VaR model encountered by the bank, is offered by Will Rhode, principal and director of fixed income research at New York’s TABB Group. “VaR fails to account for liquidity risk and the instrument they were choosing to hedge did not have enough liquidity in its space to accommodate the trades,” says Rhode.
Liquidity risk that Rhode mentions, which can be defined in various ways, can indeed be one failing of VaR models. Most critically, VaR models each institution in isolation and does not contemplate systemic events that would bring down all institutions simultaneously. VaR statistical weaknesses also include the fact that they rely upon bell shaped curves with “normal” distributions for their calculations. Technically VaR identifies the return on a specific point on the tail of this normal distribution. But in real life, returns don’t always follow this bell shaped curve. They can exhibit “skewness” (meaning returns are asymmetric) or “fat tails”, which are extreme, black swan like events. VaR may therefore not fully convey the financial risks resulting from extreme events.
Rhode offers a compelling scenario of what might have gone wrong with the model and hence J.P. Morgan’s position. Essentially he argues J.P. Morgan’s trading position became too large or just too illiquid to be able to liquidate on short notice — or for that matter for the VaR model to pick up. Rhode makes an analogy to Amaranth Advisors that lost $6 billion dollars on natural gas bets in 2006. “Liquidity risk came into play after there were no natural counterparties with whom to unwind its positions, which amounted to nearly 10 percent of the global market,” Rhode writes in a research note. He asks, “Could it be that J.P. Morgan got itself into the same situation as Amaranth Advisors?” Amaranth was an energy-trading hedge fund that lost $6 billion of its $9 billion in assets in September 2006 after its VaR model vastly underestimated its exposure to liquidation risk. In short, according to this analysis, J.P. Morgan faced an event not anticipated by VaR models, throwing the model and the bank’s trading positions out of whack.
But not everyone is ready to finger VaR as the cause of J.P. Morgan’s problems, including some very prominent critics of the risk measure. NYU Stern professor Thomas Philippon, who has identified weaknesses in the VaR approach and created measures of systemic risk not captured in VaR models, is not ready to dismiss the usefulness of VaR in this particular instance. “It is not clear that the problems are related to a systemic bet,” he says. There has been no evidence so far of a systemic meltdown, and the huge position held by J.P. Morgan is exactly the sort of risk that a conventional VaR model should capture. “Maybe it is a case of fat fingers (trader error) rather than fat tails,” adds Philippon.
Though something clearly went wrong at J.P. Morgan, fingering VaR as the culprit, is at this point, still premature.
