Chapter 13

Systemic Risk

Markets can remain irrational longer than you can remain solvent.

—J. M. Keynes

Financial markets on occasion experience abrupt and sometimes severe disruptions. As a consequence, market returns are more appropriately modeled using heavy-, or fat-, tailed distributions rather than standard Gaussian (normal), or log normal, distributions. During these episodes, risk models, such as traditional VaR, often fail to predict the duration and magnitude of extreme losses because they are parametrically ill suited to that task. Standard symmetric, two-parameter densities like the normal density do not produce the extreme outcomes of recent experience, and during periods of turmoil, these restrictions are especially costly as behavior becomes more complex as displayed by distributional asymmetries, skewness, and kurtosis (fat tails).

Systemic risks arise on the confluence of many factors and are accompanied by widespread losses and spikes in correlations across asset classes. The common factors frequently underlying market bubbles are discussed in Sullivan (2009). These include speculative leverage, investor emotions, and a misunderstanding of the true consequences of financial innovation, to name a few. The failure of Lehman Brothers, for example, announced on Sunday, September 14, 2008, signaled the severity of the collapse in credit markets, which precipitated a series of extreme losses that spread across asset classes as well as geographic boundaries. I demonstrate in this chapter the degree of non-normality of returns with particular attention to the shape of the left tail of the distribution of returns recently witnessed across various global markets. We also cover multivariate models of risk that condition on three common factors (liquidity, volatility, and default risk) giving us probabilities of instantaneous failure (hazard rates) as well as joint failure across markets (systemic risk). A failure event is defined as a threshold exceedance. Together, these models form a basis of monitoring tail risk and providing conditional forecasts of extreme returns both within and across markets.