|
Risk policies for active asset managers Authors: Dario Brandolini, Massimiliano Pallotta, Raffaele Zenti Source: Journal of Asset Management, vol. 4, n°6 Date: April 2004 |
Abstract
Risk management is a subject that is tending to be given more and more attention. Various risk indicators can be derived using an estimation of the probability distribution of portfolio returns. Risk indicators can then be used in a risk policy, in order to manage portfolios while keeping risk under control. Numerous articles are dedicated to the subject of risk indicator estimation, while little attention is paid to risk policies. In this article the authors do not consider estimation techniques, but focus instead on risk policies. They investigate whether the use of risk indicators, other than the commonly chosen tracking error, in a risk policy can improve the risk-adjusted relative performance of an actively managed portfolio of European stocks. Hence the performance of the risk policies is evaluated in the relative risk/return space. In that, their approach is different from Gupta and Kartinen's (2000) approach, which analysed some risk policies based on VaR, considering total risk and return rather than relative risk and return.
This article first describes the four different risk policies the authors have chosen to test. These risk policies are based on the following risk indicators: the tracking error (TE), the relative value at risk (ReVaR), the expected shortfall (ES) and the excess return at risk (EaR). In a second part, the authors give a description of the model used to estimate the ex-ante risk required to test the different risk policies. This model is the parallel filtered bootstrapping technique proposed by Barone-Adesi et al. (1999). It enables financial scenarios to be generated over arbitrary investment horizons using the information content of daily data about the relative or absolute risk that a portfolio might exhibit. Then the authors describe the procedure implemented to test the effectiveness of the risk policies. The testing period covers seven years from October 1995 to October 2002. During this period, 1,000 active portfolios made up of stocks from the Dow Jones Eurostoxx 50 were randomly generated each month. The ex-ante risk was estimated using the preceding two years of data and, if necessary, portfolio holdings were modified to keep their risk below the maximum amount allowed. In order to have a reference, a null policy was defined as the result of no risk control.
According to their results, only the risk policy based on EaR determines a different distribution of cumulative excess returns from the null hypothesis, as is shown by a two sample Kolmogorov-Smirnov test. In order to compare the risk policies, the authors computed a risk-adjusted relative performance index, which is the ratio of the upside deviation (UD) and the downside deviation (DD) of the distribution of the cumulative excess returns. This ratio allows for a comparison of the risk policies' capability to limit bad results while adding value to the process. They found that only the risk policy based on EaR had a UD/DD ratio that was significantly higher than one, while the others exhibit ratios that are slightly lower than one. They also noted that the risk policy based on TE presents a UD/DD ratio of 0.975, which is barely higher than that of the null policy, which was 0.967.
They conclude that the risk policy does not help to improve cumulative excess returns, but that it can influence the shape of the distribution of excess returns. However, only the risk policy based on EaR appears to be able to bring about some improvement, while the risk policy based on TE, which is very popular among asset managers, does not provide any significant improvement over a portfolio strategy without risk control.
References
Barone-Adesi G., Giannopoulos K., Vosper L., “VaR without Correlations for Non-Linear Portfolios”, Journal of Futures Markets, vol. 19, August 1999, p. 583-602.
Gupta F., Kartinen S., “Bound to Rebalance”, Risk, June 2000, p.71-74.
Related links
