It’s easier to list what’s wrong with Value at Risk than what’s right with it. It fails to reflect liquidity risk. While it is intended to show the likelihood of a loss exceeding the VaR measure, it cannot give guidance as to what the scale of that loss will be.
And it is by definition backward-looking. For the measure to be believed, one must assume that risks in the future will be largely of the same nature — even if not of the same order — as risks in the past. Plainly, this is not true, as crises tend to show.
There is also no consistency. Much has been made of how firms gradually shifted the goalposts during the crisis. Moving confidence levels from 99% to 95% neatly knocked out a pesky 4% of tail risks whose magnitude was soaring. Changing the timeframe of historical data used in calculations is another favourite ploy.
At least these aspects of methodology are disclosed. Much more damaging is the fact that transparency on precisely which risks are included within VaR and which are not is poor across the industry. Without clear knowledge of what is included, investors are incapable of drawing any comparisons across firms — a fact that some heads of trading businesses regularly point out.
A perfect example is the confusion around which VaR models JP Morgan used to risk-manage its Chief Investment Office, source of the $2bn black hole. In its May 10 conference call to reveal the trading loss, JP Morgan CEO Jamie Dimon disclosed that the bank had adopted, for the first quarter of 2012, a new methodology for calculating the CIO’s VaR.
But JP Morgan had never revealed this step before — neither in its full year report for 2011 nor, more disturbingly, in the original 8-K filing announcing its first quarter 2012 numbers on April 13.
The year-on-year comparisons offer no clue. The CIO’s VaR for Q1 2011 was originally reported to be $60m, but it was the same number even when recalculated using the new methodology adopted this year, as shown in the 8-K filing.
For the period of Q1 2012, however, the new methodology produced CIO VaR of just $67m. After the size of the loss became clear, JP Morgan restated the VaR using the old 2011 methodology, as $129m.
JP Morgan has not yet detailed the precise differences between the two methodologies. A footnote in the original Q1 2012 results announcement said: “CIO VaR includes positions, primarily in debt securities and credit products, used to manage structural risk and other risks, including interest rate, credit and mortgage risks arising from the firm’s ongoing business activities.”
The exact same wording is used in the SEC filing of May 10 that restated VaR using the 2011 methodology.
What now?
It is still unclear what the ramifications of JP Morgan’s blunder will be. It will doubtless give further ammunition to a push announced by the Basel Committee earlier this month to replace VaR as a means of calculating capital requirements for market risk.
US regulators and critics of the banking sector will seize on this episode as more evidence that a push to outlaw proprietary trading is essential — even if JP Morgan protests that it has been engaged in hedging, however inadequately.
Crises typically prompt spurts of disclosure. In the aftermath of the subprime crisis, banks fell over themselves to provide more and more granular detail of the holdings of mortgage-backed securities. A failure to be transparent in such situations quickly puts firms at a competitive disadvantage.
An event like this would be damaging whichever firm was involved. The fact that it has been found at JP Morgan, a bank rightly seen as a bulwark against the storm of the US subprime crisis, only makes the uncertainty worse. If investors cannot even rely on JP Morgan to manage risk effectively, what other illusions should be shattered?
But if banks themselves cannot be relied upon to improve transparency of VaR, regulators ought to be expected to force the issue. If they do not, trust in the banking industry will be eroded even further. It is already a scarce commodity: a failure to act now will see it disappear altogether.
More far-reachingly, JP Morgan’s trading loss hammers home again the point that risk management wizards seem to find it hardest to understand — even the best models can fail.
Reforming or replacing VaR is now overdue. But regulators must avoid falling into the trap that has caught the eggheads running CIO — and JP Morgan.
Anyone using or relying on a risk model must remind themselves every morning that it could go wrong. The financial system has to built with that expectation in mind, rather than with the assumption that the models will be right this time.