The genesis of the credit crisis has many origins but the last eighteen months have highlighted multiple weaknesses within the financial sector; in particular, the comprehensive failure of risk management to identify and mitigate extreme value risk and the dependencies between them in posing systemic risk. As events unfolded and financial institutions previously seen as "too big to fail" disappeared, the rest of the financial community was left to unravel the mess and understand its exposures with each of its counterparties across various asset classes.

Ignorance of fundamental market dynamics in risk management, of its cyclical nature and the feedback mechanisms existing between market stresses has played an important role in the current economic downturn. The procyclical nature of short-sighted calibration techniques is a case in point: while familiar words such as ‘bullish' and ‘bearish' are commonplace in the financial pages, few models explicitly represent such ‘states' of the market. Instead current market conditions are reflected in model parameters by only calibrating to the most recent data. As past states beyond recent history are not incorporated into model predictions, these techniques miss the bigger picture, encouraging unmeasured and excessive risk taking in the good times and over-reactive risk aversion in the bad.

A further subject of post-crunch criticism is that stress testing and scenario planning prior to the crisis failed to prevent its occurrence and therefore could not have been fit for purpose. Nonetheless, it seems unfair to dismiss the technique or the technician given the complexities of globalisation and the deluge of available information in the market. Instead one would do well to augment and inform testing with systematic processes suitable to the task. In this regard state-based pattern analysis methods can provide a natural view of market stresses that are not influenced by expectation or personal bias nor by the limits of human comprehension to assimilate high dimensional data.

Recent research by Misys on the propagation of extreme risks has demonstrated a wide range of potential applications in the banking arena that could address significant problems highlighted during the crisis. As with stress testing the curse of dimensionality presents challenges unique to any study of market level dynamics. However, use of pattern analysis algorithms enables a market wide view of risk that can reveal previously undisclosed concentration risks and evaluate overall exposure across a whole enterprise.

Application of this approach to the popular Credit Risk+ model not only has the ability to represent extreme risks but also presents a flexible means to tackle the problem of sector (and hence default) dependence, at the heart of a problem that needs to be better understood, and ultimately better managed. Moreover, institutions testing business strategies with earning forecasts that don't incorporate such factors into credit loss models miss the opportunity to identify key exposures and refine their strategies based on best practice.

Perhaps the most striking and publicly embarrassing aspect of recent crisis was the revelation that banks had failed to retain sufficient capital to operate in adverse conditions. Without adequate provisions many institutions have been forced to rely upon government rescue packages, leaving them ever-more vulnerable to political and reputational risk and raising the spectre of unwelcome and unwieldy intervention in the future.

The relevance and need for accurate and robust measures of extreme value risk and the dependencies that exist between them has never been more apparent. Examining patterns in market movements rather than an exact mathematical analysis will help to deliver a snapshot of the situation at any given time. Armed with this insight, it is far easier to mitigate risk exposure across each of the business lines throughout an organisation. BT

 Yimin Liu and Noel McWilliam are financial engineers at Misys Risk