According to this Fundspeople article as you will find summarised below, volatility as a moderately reliable tool for managing investment portfolios is fading in the face of the new financial normal in which we live. Inaccurate interval averaging and uncontrolled shifting correlations (or controlled by political decisions and central bank interventionism as never seen before) mean that much greater errors are being made in calculating portfolio volatility today than just 5 or 10 years ago. The term that defines this financial unpredictability is heteroscedasticity, and it is here to stay. Let's look at it:
The traditional approach to portfolio volatility is inherently limited. It is common for investors to measure historical volatility by looking at the standard deviation of interval returns over a defined period. By combining this information with historical levels of correlation between different asset classes or securities, investors seek to diversify while distancing their portfolios from irregular risk (diversifiable risk). Typically, investors do this by diversifying their exposures across multiple asset classes, sectors and regions. The modern portfolio theory approach is only effective if certain conditions are met.
According to a study carried out by the asset management firm Pioneers, The first refers to standard deviation/kurtosis. For the standard deviation of returns to be an accurate measure of volatility, the return intervals must be normally distributed and independent. In most cases, the validity of these claims cannot be determined and, although the standard deviation of the return intervals is relatively easy to calculate, it will only provide an approximate representation of the level of risk (sic), according to the organisation. By way of illustration, consider the distributions of the daily returns of the MSCI World Index over the last five years (see Figure 1).
_09.46.12.png?1435912711)
The figure shows two main differences between the normal distribution (orange curve) and actual returns (blue bars). Firstly, Real returns show a greater number of events around the mean or median return, and their peaks are higher. Secondly, real returns exhibit ‘fatter tails’ than returns with a normal distribution. Excess kurtosis refers to the probability that ‘extreme events’ will occur more frequently than expected — which has significant implications. For example, over the last five years, the daily return on the MSCI World Index has fallen by -3% or more on four occasions, which means this has occurred 25 times more frequently than would be expected under a normal distribution of returns; in other words, extreme events occur more frequently than expected.
The second reason relates to correlation dynamics. To optimise a portfolio so that it actually experiences a lower level of volatility, the correlations between assets must remain stable over time. However, Pioneer points out that recent episodes of turmoil in the financial markets have posed challenges for traditional asset allocation strategies. Over the last eight years, during periods of severe market stress, portfolios diversified along traditional lines (by asset class) suffered substantial losses, as many asset classes fell simultaneously, despite their historical performance showing low correlations. These losses were not solely caused by increased volatility across all asset classes, but also by changes in correlation dynamics, the firm points out.
In this regard, Correlations between different asset classes have increased significantly since 2007. Figure 2 shows the trend in the average correlation between different asset classes over the last twenty years and the CBOE Volatility Index (VIX). Correlations had not shown a downward trend until recent years, given the environment of low volatility created by the accommodative monetary policies of central banks around the world. This trend appears to have changed during the fourth quarter of 2014, even though the VIX index remains at low levels. “We believe that, should the volatility index rise, there is a real risk that correlations between asset classes will increase even further,” they state.
_10.00.35.png?1435912867)
There is volatility in the volatility of portfolio returns. When relying on historical volatility as a measure of risk, investors may not realise that the current level of portfolio volatility deviates from its historical level, meaning they may be unaware of the true amount of risk they are taking on. We know that correlations are not stable over time and that the variance of return intervals is not constant over time — they are said to exhibit heteroscedasticity. Heteroscedasticity refers to the situation in which the variability of one variable is uneven across the range of values of a second variable that predicts it. “It is an important concept because the standard deviation of returns is not constant, but tends to fluctuate depending on the time period analysed. This means that investors run the risk of underestimating the actual levels of risk in their portfolios”, they point out.
Source: FundsPeople