Reinforcing the conclusions of a white paper sent to clients in November, Wilshire Analytics’ global equity risk model, the Wilshire GR6 Equity Risk ModelSM, shows the year-end spreads between the model’s long-term and short-term valuations of risk are at record highs.
Global average of equity risk spreads between equally-weighted and exponentially-weighted risk
November 2008 December 2007 Previous Record Spread (July 2002)
10.54% 1.28% 2.57%
The paper, “Wilshire Equity Analytics: A Comparative Study of the Wilshire GR6 Covariance Matrices,” by Edward Rackham, PhD, vice president, Wilshire Associates, shows the impact on portfolio risk of five different risk estimation schemes.
“Dr. Rackham’s research shows that in order to be useful within the investment process, a selected risk model needs to reflect the volatility of the expected trading horizon,” commented David L. Hall, senior managing director, Wilshire Associates and head of Wilshire Equity Analytics, a division of Wilshire Analytics. “A manager using a model with a time horizon inconsistent with the investment horizon means that the manager may be managing to the wrong estimates of portfolio total risk and tracking error.”
Dr. Rackham, who oversees the development of new risk, performance and other analytics functionality in the Wilshire AtlasSM, the premier solution offered by Wilshire Equity Analytics, commented that the methodology used to estimate covariance can lead to significantly different estimations of portfolio risk.
“The paper also discusses how the Wilshire Equity GR6 Risk Model’s 2005 implementation of Wilshire’s proprietary Structured Hadamard Product Target Shrinkage Estimator (SHaPTSE) improves the performance of the model’s short term risk estimates through ‘shrinkage’ of any spurious ‘off-diagonal elements’ found in the model’s daily covariance matrices,” Dr. Rackham said. “As I noted in the paper, “…in comparison to longer-term risk estimates, “it seems reasonable to conclude that the best one-month forward-looking predictions of risk … are yielded by the Daily Exponential … matrices.”
Dr. Rackham cautioned that this does not mean that long term risk practitioners should stop using monthly models as they may prefer more stable risk numbers rather than concern themselves with sensitivity to daily fluctuations. “Users of risk analytics need to be aware that in times of extreme market volatility differences between the risk estimates of different models tend to be exaggerated,” he explained.