Extreme Value Theory and stress testing, in combination with factor-based risk models, can help investors around the shortcomings of VaR, says Jennifer Bender

As we move further into 2009, the ongoing shakeout in the financial sector and recessionary woes continue to fuel uncertainty in the global equity markets. Beginning in September 2008, major stock market indices declined dramatically, adding to the year’s already dismal performance, and other markets such as corporate bonds and commodities have also experienced dramatic declines.

Historically, volatility has risen when asset prices have declined. This phenomenon has many possible explanations but, in general, when equities have declined, investor nervousness has increased and this has translated into greater uncertainty across the market. Market fluctuations thus have tended to become more pronounced in bear markets. In autumn 2008, days when major equity market indices bounced up and down by more than 3% became commonplace. This never happened on any day in 2007.

What is unusual about this period is the strength and rapidness with which risk has increased. Risk has far surpassed its previous 2002 peak. Relative to the past 20 years, the recent increase in annualised volatility of the MSCI All Countries World index has been unprecedented, with October 2008 experiencing the largest month-over-month increase. The prior record had occurred in August 1998.

What stands out in this episode is the far-reaching impact on the bear market. Confirming this anecdotal evidence, we see that, in fact, systematic risk, relative to specific risk, has soared in importance. Using BarraOne, our global multi-asset-class portfolio risk and performance attribution platform, we are able to analyse the risk of the largest 2,000 or so stocks in the European equity market, as illustrated in the figure below right. We show the portion of risk (averaged across stocks) that is systematic versus specific (or idiosyncratic). The proportion that is systematic - risk that is ‘market-wide’ and not specific to certain groups of stocks - has skyrocketed in recent months. This rise in systematic risk means that the diversification benefits of investing in a large number of equities have declined dramatically and, consistent with the experience of many well-diversified institutional investors, even small amounts of market exposure have been extremely painful this past year. 

Naturally, one of the most pressing questions for institutional investors is how best to manage risk in these times. Of course, with 20/20 hindsight, the best decision may have been to park all assets in cash. Unfortunately, timing bubble peaks and corrections is a bit like forecasting hurricane paths. It can only be done sometimes and usually only a day or two before the hurricane hits. Even if investors know an event like the 1998 Russian rouble default will happen with certainty, predicting when is a Herculean task.

However, the very fact that markets will likely continue to exhibit unpredictable, and sometimes irrational, fluctuations opens opportunities to investors who institute sound risk management. For traditional pension plans a certain degree of market exposure is unavoidable and performance will necessarily be affected by asset class cycles - particularly if, like in 2008, the troughs in these cycles coincide. Through good risk management they can understand what risks they are exposed to and to what extent, well in advance of financial storms, and make sure they have enough liquidity to weather them.

There is a breadth of risk management tools available to institutional investors these days. Value at Risk (VaR), which has been for many years the regulatory standard, has come under fire in recent months. VaR captures the point at which investments go sour, but doesn’t measure how much the investor actually loses if the worst occurs. Moreover, most standard methods used to calculate VaR rely on assumptions that are unrealistic when markets turn. For instance, returns are often assumed to be normal even though it is actually the non-normal ‘tail’ events that investors are most concerned about.

Another common practice is to use historical returns to estimate VaR - which clearly overlooks potential events that have not occurred before. Similar measures like Conditional VaR, also known as Expected Shortfall (which captures what investors are likely to lose on average when the VaR limit has been exceeded) are also susceptible to these issues. To address the current limitations, risk researchers have more recently focused on statistical methods like copula estimation and Extreme Value Theory, which can at least account for the presence of tail events in asset returns. These solutions offer promising avenues for resolving the p erceived shortcomings of current VaR estimation techniques.

Many other tools are also available and in many cases are under-utilised. Stress testing is a framework in which investors shock portfolios with either past events or hypothetical events (for example, a 10% market drop or a 50bps interest rate hike) to assess how their portfolios would react. Stress testing is well suited to extreme event analysis and allows institutional investors to consider new and historically-unprecedented combinations of events (such as the rapid drop in equity and oil prices we witnessed in September 2008). Stress testing also allows institutional investors to shock certain sources of risk like systematic factors. In combination with a factor-based risk model, institutional investors can quantify the potential risk of a wide range of factor shocks, like market events originating in certain countries or industries.

For many institutional investors, this past episode may be causing a complete overhaul of their long-range objectives and the amount of risk they want to take. Before the crisis, five years of low volatility had spurred a greater push towards higher-return and higher-risk assets, hedge funds, commodities, real estate, private equity - any source of alpha which could both enhance returns and provide diversification. In this market, we may see institutional investors seeking lower-volatility alternatives where they can satisfy their need for equity exposure while reducing risk.

Minimum-volatility, managed-risk, and low-  beta equity strategies are seeing increasing interest. Other investment tools such as portfolio insurance and the use of derivatives to hedge losses in down markets may also attract attention, particularly if the current market downturn persists. For instance, buying equities and selling out-of-the-money calls on the equities generates an option premium that can buffer loss in down markets.
Or for even more risk-averse investors, buying puts to protect losses, while expensive, can offer the best downside protection. These and other option-based strategies are likely to become more appealing if the current high-risk climate continues.

Most importantly, investors should continue to have ‘long memories’ even after the market resumes its growth. In low-volatility periods like 2003-2006, active returns become even harder to generate, causing portfolio managers to take increasingly riskier bets to meet their return targets.

Moreover, the emphasis on risk management at many institutions often gets down-played. For asset owners, understanding the variations in volatility cycles and setting appropriate return and risk objectives for asset managers may be the first step in ensuring that they are not primed for a severe fall when the market crisis hits. Putting safeguards in place for when risk starts to rise is also important component. Ultimately, it is the focus investors place on risk management when volatility is low and markets are smooth that determines how they will fare during the next round of market turmoil.

Jennifer Bender is vice-president in applied research with MSCI Barra