A typical market conditions such as todays shine a spotlight on risk measurement
Calm before a storm?
To see how atypical current market conditions are, consider for example equity market volatility. This has been exceptionally low: the volatility of daily returns over the past six months has been just 7.1%1, which is less than half of the longer-term average. The last time that volatility was this low was late 2006/early 2007 – and you don’t need reminding of what happened next on that occasion. Forward-looking measures tell the same story: the widely-tracked VIX2 index closed last Friday at 9.61, which is also well below historical norms.
It’s not just the equity market; it is a similar story across other asset classes from fixed income to real estate to commodities.
It’s not just volatility: correlations are also unusually low. The average pairwise correlation of U.S. stocks is close to historic lows, and cross-sector correlations are also well below normal levels. And just as volatility regimes can change, so too can correlations. Indeed, Stas Melnikov3 – who knows about these things – tells us that correlations are much more difficult to forecast than volatility.
Given that market valuations are high4 and there’s plenty of geopolitical and economic uncertainty, this low-volatility-low-correlation regime might seem to be vulnerable. A closer look at volatility markets supports this perception. The VIX, for example, is based only on the outlook for the next thirty days so for a longer-term outlook we need to look at the term structure of VIX futures prices; that term structure is currently steep, implying higher volatility the further out one looks. Similarly, the SKEW measure (a measure of tail risk) has been well above historical averages in the past few months.
So there’s good cause to believe that even though the prevailing market conditions seem benign, the situation is more fragile than it might seem.
Because investment risk is a complex thing, it needs to be measured from multiple perspectives, as illustrated in the diagram above. "Distributional analysis" is the starting point and this refers to the output of risk models that project distributional measures, such as volatility or tracking error, and potential losses such as VaR or CVaR5. These models tend to base their assumed return distributions on historical experience. So, because market conditions can change, these results should be supplemented with other analysis.
Sensitivity analysis varies input assumptions to test the robustness of a model’s results: we might, for example, substitute typical (or "flat") volatility and correlation assumptions for the prevailing market conditions. Sensitivity analysis is used to obtain insight into the sources of risk exposures: what are the ultimate drivers of risk for a particular investor? The answer is different for a pension plan than it is for a For-Purpose organisation or a sovereign wealth fund or an individual investor.
Scenario analysis looks at what might happen if the prevailing environment changes for the key variables. It brings out, for example, the critical role of correlations. If equity markets were to drop, the impact on investors would be quite different depending on whether that’s accompanied by falling or rising interest rates. The falling rate scenario would be bad news for pension plans, whereas for For-Purpose organisations or individuals, it’s rising rates that are likely to be damaging (because of the implications for inflation.) Stas emphasises that scenario analysis involves a heavy dose of art to supplement the science.
The fourth component of the framework above is back-testing, comparing forecasted outcomes to actual returns. Models are only useful if they reflect reality, so we need to measure how well the models are performing.
These tools provide the information investors need if they are to be prepared to respond appropriately if and when prevailing market conditions change. In such a market, the more perspectives you’re using to measure the risks that matter to you, the better. The multi-pronged approach is, we believe, best practice in all market environments. But when prevailing conditions are as atypical as today’s environment, it becomes all the more important.
1 Annualised volatility of S&P 500 returns for the six months to October 13, 2017.
2 CBOE volatility index.
3 Stas is Director, Investment Risk at Russell Investments and heads the team responsible for oversight of investment risk.
4 The U.S. market cyclically adjusted price-to-earnings (CAPE) ratio is currently above 30, for example, well above typical historical values.
5 VaR is value at risk and typically represents the maximum expected losses that might be incurred over a given time horizon with (for example) a 95% confidence level. CVaR is conditional value at risk and refers to the average loss expected to be incurred within the 5% tail. Both measures are highly dependent on the assumed distribution of returns.