The Truth About Tail Risk

Coverage of the 2011 Global Investment Conference

  • Facebook
  • Twitter
  • Print
  • Email
  • Comment

507311_whale_tail_series_4The global financial crisis provided a stark example of the fact that risk and correlation parameters can be wildly unstable through time. But it was by no means the first example. From the stock market crash of 1987 to the Russian debt default in 1998 to the “tech wreck” in 2000, sophisticated investors recognize that economic and financial conditions are subject to abrupt regime shifts. This instability presents significant challenges to portfolio construction and risk management. In this article, I will discuss tools that investors can use to stress test their portfolios, monitor the likelihood of tail risk, and dynamically reallocate in anticipation of regime shifts. I will also draw on research conducted in collaboration with Mark Kritzman, CEO of Windham Capital Management and a professor of financial engineering at MIT’s Sloan School of Management.

Turbulent markets are characterized by correlation break-downs, not just volatility

Investors have access to a wide range of volatility-based risk measures to help them forecast volatility and performance. Correlations between investments, on the other hand, are often overlooked as a predictor of risk-adjusted returns. To measure market turbulence, we rely on a statistical measure known as the “Mahalanobis distance.” Without getting too technical, the measure can be thought of as a multivariate z-score.[1] It describes behavior across two or more assets and it detects both heightened volatility as well as unusual correlation outcomes. It enables us to construct a single, daily measure of turbulence for a particular market. Or, we can apply it to a specific portfolio or roster of active managers.

Why doesn’t volatility tell the whole story?  Imagine a situation where two industrial sectors, which typically move in near lock step, suddenly diverge. One sector (say, telecom) experiences a significant gain while the other (perhaps info tech) plummets. Furthermore, assume that the moves are perfectly offsetting. In other words, the return of the overall market is zero and its volatility is flat. But beneath the surface, this was a highly turbulent day. By accounting for correlation effects, turbulence gives us a more comprehensive picture of risk.

This measure of financial turbulence has several other attractive features. First, it is intuitive; it tends to spike during recognizable market crises and events. Second, there is a strong link between the level of turbulence and the performance of key asset classes and investment strategies. On average, returns to risk are substantially lower during and after turbulent episodes. Third, turbulence is persistent, which means it typically has an adequate shelf life as an alpha signal. Taken together, the second and third points mean that investors can enhance returns by de-risking when turbulence first strikes.

Monitoring systemic risk by measuring implicit links between companies

The crisis also provided stimulus for investors, regulators, and academics to put the concept of systemic risk under the microscope. Some of these initiatives have focused on building complex databases to monitor the extent to which firms are linked together by, for example, derivatives contracts. When firms are closely linked together, the market is more susceptible to systemic failure if one of the dominoes topples over.

While we applaud the focus on systemic risk, we believe that the network of relationships between firms is so complex that a direct attack on the problem is impossible. Instead, we have focused on measurements of systemic risk that can be implied from securities prices.[2] Specifically, we compute the absorption ratio, which equals the fraction of variation within the market that is explained (or absorbed) by a small number of macro risk factors. To identify the factors, we use a statistical technique called Principal Components Analysis.

When the proportion of variation absorbed by a small number of factors is increasing, we conclude that systemic risk is high. In these environments, markets are largely “risk on/risk off” and are more fragile; shocks may propagate quickly and broadly and result in large drawdowns. On the other hand, when the proportion of variation absorbed by a small number of risk factors is decreasing, we conclude that markets are less systemic and that stocks are responding to more industry- and company-specific news.

How can investors apply these concepts in the real world?

We see three applications for measures of turbulence and systemic risk:

  1. The measures provide a useful basis for stress testing. By examining how a particular manager, asset class, or portfolio has performed during different environments, an investor can appropriately set expectations.
  2. Investors can monitor the level of turbulence and systemic risk on a live basis, and in some cases receive advance warning of tail risk events. For example, our research has shown that turbulent periods are almost always preceded by rising systemic risk.
  3. Because returns to risk – such as the equity risk premium, the small cap premium, and the carry trade – are lower during periods of high turbulence and systemic risk, investors can enhance performance by de-risking when these measures spike.

Will Kinlaw is managing director and  head of the portfolio and risk management group at State Street Associates.

[1] See: Kritzman, M. and Y. Li.  “Skulls, Financial Turbulence, and Risk Management.”  The Financial Analysts Journal, May/June 2010.

[2] See: Kritzman, M., Y. Li, S. Page and R. Rigobon.  “Principal Components as a Measure of Systemic Risk.”  Revere Street Working Paper Series: Financial Economics 272-28, updated June 21, 2010.

Add a Comment

Have your say on this topic! Comments that are thought to be disrespectful or offensive may be removed by our Canadian Investment Review admins. Thanks!

Transcontinental Media G.P.