Because of its importance in the current regulatory framework, the evaluation of value-at-risk (VaR) continues to receive considerable research attention. This issue of The Journal of Risk contains papers that propose a number of different possibilities for improving the estimation of VaR or for complementing its interpretation, thereby offering enhanced tools for risk assessment.
As illustrated very starkly by the major financial events of recent years, liquidity plays a significant role in times of crisis. The paper "Accounting for nonnormality in liquidity risk" by Cornelia Ernst, Sebastian Stange and Christoph Kaserer contains an improvement on the standard approach that relies on the normal distribution to incorporate liquidity risk into VaR estimation. A new and easily implementable parametric approach based on the Cornish-Fisher approximation is proposed in order to account for nonnormality in liquidity risk. The authors demonstrate how to implement this methodology for a large sample of stocks and they provide evidence of their methodology's accuracy.
Typically, VaR is first estimated on a daily basis and then extrapolated to longer horizons by scaling it by the square root of the length of the longer period. This practice is valid only for daily returns that are assumed not to be serially correlated. However, this assumption does not hold for real data and the practice may result in misleading values. In the second paper in this issue, "Scaling portfolio volatility and calculating risk contributions in the presence of serial cross-correlations", Nikolaus Rab and Richard Warnung develop an approach that accounts not only for time series correlation but also for cross-sectional dependence. The authors show that ignoring such correlations can lead to significant volatility bias as well as incorrect risk attribution in portfolio evaluation.
While VaR is an indicative measure of risk, its definition as a quantile means it does not offer anything about what occurs specifically in the remainder of the tail distribution. It has therefore been criticized for encouraging excessive but remote risk. The paper "Value-at-risk and ruin probability" by Jiandong Ren contains an analysis that illustrates how ultimate ruin probability may actually provide useful information about financial or actuarial risk, thereby demonstrating its complementary role in addressing some of the criticism of VaR.
The tail distribution is critical to the evaluation of VaR. However, rare events and limited data make its empirical estimation very challenging. In the paper "Fitting the generalized Pareto distribution to commercial fire loss severity: evidence from Taiwan", Wo-Chiang Lee proposes the use of fire loss data to help better evaluate different common parametric approaches for data-driven tail loss distributions.
Scaling portfolio volatility and calculating risk contributions in the presence of serial cross-correlations