Canadian philosopher Marshall McLuhan once said that “we drive into the future using only our rear-view mirror”. Nowhere is this description more accurate than in the investment industry and specifically in the area of risk management
Canadian philosopher Marshall McLuhan once said that “we drive into the future using only our rear-view mirror” nowhere is this description more accurate than in investment industry and specifically the area of risk management.
Rather than engaging in a realistic and informed conversation about future risks, the industry has become myopically focused on a simplistic backward looking measure - volatility. In this short article we examine the problems inherent with this simplistic approach, and suggest steps to improve the accuracy of the risk profiling process to reduce the suitability mismatch that we perceive with some existing tools.
The adoption of simple risk mapping tools that relay on volatility is an understandable reaction to increased regulatory scrutiny and the disappointing performance of many multi-asset portfolios during the financial crisis. However, in an attempt to create a demonstrable process for matching clients to funds, risk assessment tools have tended to ‘boil down’ risk to the single measure of volatility. In doing so, these tools ignore other risks including: permanent capital loss, illiquidity, default, failure to meet objectives, and counterparty.
While the use of volatility as a proxy for risk provides a statistical basis for describing the randomness of capital market movements, its reliance on assumptions and its poor predictive power mean that volatility is both a weak proxy for risk, and an unreliable way to predict severe capital loss. It is therefore of limited use in matching funds to clients.
The calculation of volatility makes two big assumptions: first, that returns are normally distributed, and second, that correlations are stable. Neither is true. A cursory glance at equity return data over very long periods shows that the distribution of returns is subject to both skewness and positive kurtosis. This means that the typically used metrics of mean return and volatility do not fully describe the distribution of returns.
The problem is magnified when using historic data from multi-asset portfolios to predict the future. Traditional risk tools tend to use average correlation data. In reality the correlation relationships that create the diversification benefit of a multi-asset portfolio are unstable. Correlations tend to increase during periods of stress, reducing diversification benefits and increasing losses beyond that which would be expected using average data.
As a consequence of these failings, investors who base their risk assessment on volatility are like McLuhan’s driver, they are susceptible to nasty surprises that are not obvious from the view they have of the market. By focusing on absolute levels of volatility as the key measure of risk, investors are prevented from buying risk assets when prices are low as these typically corresponded to periods of high volatility. Equally, portfolio managers are encouraged to buy risk assets when prices are high. This buy high, sell low strategy is unlikely to be in the clients’ best interests.
The practical problems with this approach are especially evident when using absolute levels of volatility to match funds to client risk profiles. Morningstar has recently conducted research that shows that the volatility of a conventional multi-asset portfolio varies widely through the market cycle. We created a series of multi asset portfolios and tracked their volatility using the approach stipulated for the calculation of a fund’s synthetic risk return indicator (SRRI) that is included in key information documents (KIID). The volatility of these portfolios varied significantly over time. For example, the volatility of a moderate risk portfolio comprised of recognised benchmark indices varied by 5.3% over the last 9.5 years. This volatility range is greater than the SRRI band (four) used to classify the fund. This means that a portfolio positioned in the middle of an SRRI band at the beginning of the period and rebalanced regularly would breach both the upper and lower boundaries of that band over the period. In other words, without changing the allocation, the portfolio fund would be both too risky and not risky enough for the same client over the period. A risk mapping process that produces such widely varying results for a stable portfolio is clearly not fit for purpose.
To overcome these problems advisers and investors need to spend less time looking in the rear-view mirror and instead focus on their instruments and the view through the windscreen. In practical terms, there are three ways this can be achieved.
The first is to improve the quantitative the tools used to assess risk. As volatility has become increasingly discredited, many investors are moving towards more sophisticated measures such as the ‘mean conditional value at risk’ that focus on the potential loss in an extreme event. Second, risk must once again become a conversation between the adviser and client rather than a simplistic ‘tick-box’ exercise. Third, as an industry we need to improve both our qualitative methods of assessing the risk of a fund. The objective, attitude, and investment process of the manager is at least as important as the past performance of the portfolio. Managers with similar starting portfolios may generate entirely different results through a crisis, depending on the temperament, past experience and mandate of the investor. It is only through an understanding of the manager that a fund buyer will be able to discern the likely differential in the performance of superficially similar funds across a range of market conditions.
This greater focus on the future rather than the past should considerably improve the changes of the investor reaching their destination safely.
By Dan Kemp, Chief Investment Officer, EMEA