| MSCIBarra Full Review |
|
Investment Risk The riskiness of an investment is related to the certainty of its outcome. The less certain the outcome, the riskier the investment. For example, the range of possible returns on a share investment is wider than that on a bond investment, and shares are therefore considered to be riskier than bonds. The riskiness of an investment can thus be equated to the volatility of returns. We talk of historic volatility (the range of returns which were realised historically), or expected volatility if we are looking to the future. Standard deviation is a useful and widely accepted measure of the volatility of returns. If we can estimate the standard deviation of returns on an asset, we can, on the basis of some very reasonable assumptions, calculate the probable range in which the return on the asset will fall. Similarly we can calculate the probability that the return will be less than a required percentage. Beta is another popular measure of risk. It measures the sensitivity of the return on a share to the return on the market (as measured, for instance, by the return on the All Share index). It reflects the exposure of a share to market risk. A share with a beta larger than one is more volatile than the market and therefore riskier than the market. Such a share is likely to rise faster than the market during a market upturn and will fall faster than the market during a market decline. Predicting Risk A common approach to estimating risk is to calculate the standard deviation of monthly returns over some historic period and to use this “realised” risk as a prediction of future risk. There is no generally best historic period over which to calculate realised risk, although 36 months (which is “neither too long or too short”) appears to be the norm. There are disadvantages to using historic volatility as a predictor of future risk. The first, and most obvious, is that unless a share or portfolio has been in existence over the historic estimation period, no risk prediction can be made. Secondly, shares and portfolios change their characteristics over time and it is really quite unlikely that price movements three years ago will have any bearing on current and future price movements. This is a particularly important consideration when estimating the correlation between share returns, which determines the extent to which individual share volatilities are reinforced or diversified away within a portfolio. A Multiple Factor Risk Model The problems in using historic volatilities of share returns in predicting future volatility can be overcome by developing and using a multiple factor model of returns in the equity market. Simply put, a multiple factor model of share return is an equation relating the return on a share to the returns on a number of factors that drive share returns in a particular market. The volatility of the share return will then be a function of the volatilities of the returns on the factors in the model. The following simple example will illustrate this approach. Let us assume that two factors influence returns on all South African shares, namely, the size of the company and whether it is a mining or a non-mining company. Using this simple two-factor model the volatility of return on a large mining company share will be determined by the volatility of returns on mining shares, the volatility of returns on large company shares, and the extent to which these two factor returns are correlated. (The two factors in this model are called common factors, ie, they affect returns on all the shares in a market. Returns can also be influenced by company specific factors, eg, ore reserves of a mining company, and the risk introduced by this specific factor – specific risk – must also be estimated to obtain an estimate of the total risk of a share). This simple example shows some of the advantages of using a multiple factor model: The MSCIBarra Multiple Factor Risk Model MSCIBarra developed a multiple factor model for the South African equity market during 1995. MSCIBarra was assisted by a consortium of 13 South African fund management organisations to ensure that the factors incorporated into the model were appropriate for the local market. Potential factors for inclusion into the model were selected on the basis of the fund managers’ intuitive and empirical knowledge of the factors driving local share prices. The final selection of factors was based on exhaustive statistical testing to determine which of the factors do, in fact, have statistically significant relationships with share returns. Factors in the South African Equity Model The factors in the South African risk model fall into two broad categories: industry factors and style factors. The industry factors are the 41 JSE industry sectors. Shares within a sector tend to be influenced by the same set of macro-economic events, and it therefore appears reasonable to differentiate between the industry classifications of shares when specifying factors that influence their riskiness. Style factors are based on fundamental company as well as market related characteristics that have been identified as making significant contributions to the volatility of share returns. The following are the essential characteristics of the twelve style factors in the South African model:
Thus the South African multiple factor model consists of a total of 53 so-called “common” factors. These are the factors that to a larger or lesser degree influence the volatility of all the shares on the JSE and account, on average, for some 50% of the volatility of individual shares returns and well over 95% of the volatility of well diversified portfolios. The risk of individual shares that is not due to exposure to these common factors (ie, share specific risk) is estimated using statistical forecasting techniques and is added to common factor risk in order to arrive at a prediction of total risk. Updating the Model, Estimation of Factor Returns and Risks Estimating risk with the multiple factor model requires estimates of factor returns, the volatilities of factor returns and the correlations between factor returns. These estimates are obtained monthly when the model is updated. Every month MSCIBarra measures the exposure of all JSE listed shares to the factors in the model. Using advanced statistical techniques the end of month returns on the shares are then attributed to the beginning of the month factor exposures of the shares. This process yields the monthly factor returns. From the history of monthly factor returns the volatilities and correlations of these returns are calculated. In these calculations more recent factor returns are given more weight in order to better reflect recent volatility trends in the market. Once these updates have been completed it is a simple process to estimate the risks of shares and share betas, which are both dependent on the factor return volatilities and their correlations. Model Testing During the monthly update the model is also subjected to a number of tests to determine how well it explains the observed returns in the market and to check the accuracy of its predictions of the risk of individual shares and that of a selected number of portfolios. These tests consistently confirm that the MSCIBarra model predicts risk accurately, that these risk predictions are superior to those based on historic return volatilities, and that these predictions are quick to respond to changing characteristics of shares and changing market volatility. Total Risks and Betas The total risk and beta of shares that are published in this book were estimated with the MSCIBarra model at the most recent quarter end. These estimates therefore reflect the exposure of the shares to the factors in the MSCIBarra model at that time as well as recent market volatility. Subsequent changes in a company’s characteristics and major shifts in market volatility can therefore affect the quality of these estimates. The total risk is an annualised number, ie, it is the predicted standard deviation of a share’s annual return volatility. The monthly equivalent can be found by dividing the annual risk by 3.46. Beta gives the sensitivity of the share return to that of the market return (as represented by the return on the All Share index). Shares with a beta larger than one will tend to outperform the market during a bull run and will under perform during a bear run, and vice versa for a share with a low beta. It is, however, important to keep in mind that factors, other than the market, influence the return on a share and share returns will not always behave as suggested by their estimated betas. Historic volatilities and betas will tend to differ from the MSCIBarra estimates. The MSCIBarra estimates are, however, more sensitive to the latest market movements and also reflect the current exposure of shares to the MSCIBarra industry and style factors that to a large degree determine share volatility. It is for these reasons that the MSCIBarra estimates are better predictors of future volatility and betas than the historic values for these parameters. Differences between the MSCIBarra estimates and the historic values will also indicate changes in the levels of riskiness of shares (as measured by the volatilities and betas). For more information on MSCIBarra’s suite of risk management products, please call Hendrik du Plessis on 021-683-3245. |
|
|