By Anindya Bhattacharyya
Jon Danielsson, the London School of Economics academic, has defended his forthright criticisms of the Basel Committee’s Value-at-Risk (VaR) based approach to financial markets regulation. Speaking last month at a round table discussion organised by the Centre for the Study of Financial Innovation, Danielsson outlined the anti-VaR thesis he put forth in a paper published last year, provocatively entitled, The Emperor Has No Clothes: Limits to Risk Modelling.
In that paper, Danielsson argued that capital adequacy regulations based on VaR models were of little use in defending against market crises and could well increase systemic risk. He urged regulators to rethink the near-unanimous consensus around VaR-based regulation and instead consider newer risk management techniques, such as those drawn from attempts to directly model market behaviour during crisis situations.
At the CSFI round table, Danielsson noted that his views, though apparently controversial, had elicited a by-and-large positive reaction from the financial industry. Traders and risk managers were generally aware of VaR’s failings, he said, and had in many cases already moved away from such measures in their internal risk management practice. Moreover, industry practitioners had made their concerns known to regulators, he added, but the Basel Committee had nevertheless chosen to continue down the VaR road in its latest proposals.
Danielsson’s critique focused on the regulatory definition of VaR, which, he claimed, rendered the metric meaningless. “VaR is the minimum loss on a bad day and the maximum loss on a good day”, he said. But regulations stipulate that a “bad day” is something occuring once every 100 days. As a result of this, capital adequacy calculations based on VaR were incapable of taking into account the probability of a market crisis or the actual size of expected losses on a “bad day”.
He added that VaR was “easily manipulated”, in that traders could structure their portfolios in order to minimise VaR by lowering profits and increasing downside risk. Others at the discussion concurred, arguing that VaR regulations encouraged traders to adopt “Martingale” strategies, ie double-or-quits bets that lower the minimum losses on a “bad day” while significantly increasing expected losses.
Danielsson also noted that the profusion of different data sets, different VaR models and different time horizons meant that the VaR figure for a set portfolio could vary by a factor of five, depending on precisely which configuration was chosen. He presented the results of one such empirical study which suggested – contrary to regulatory wisdom – that using longer time horizons for VaR calculation led to more effective risk management. On a more fundamental level, Danielsson cast doubt on whether the regulatory goal of encouraging best-practice risk management within financial institutions was necessarily compatible with the goal of ensuring systemic integrity within financial markets.
He cited the market turbulence triggered by Russia’s 1998 bond default as an example of how sound hedging practice by individual institutions can exacerbate a systemic crisis rather than abate it. By encouraging such practice through capital adequacy regulations, regulators were “harmonising preferences” and thus inadvertently increasing the severity of market crises, he said. Conversely, he noted evidence that suggests the presence of unregulated financial institutions can actually dampen market crises.
A lively discussion followed Danielsson’s presentation, with contributions from bankers, consultants, regulators and academics. Though some questioned certain technical aspects of Danielsson’s findings, most agreed that his criticisms were essentially sound. However, participants also noted that there was no simple way of addressing the shortcomings he had highlighted.
Danielsson acknowledged the complexity of the issue, but suggested that regulators could at the very least rethink their strategies by focusing on “the one bad day in a thousand rather than the one bad day in a hundred”. Specifically, he urged the use of longer time horizons in calculating risk measures and a shift from
VaR models to expected shortfall metrics that gauged the probability and likely severity of market crises.