Credit spread shocks: how big and how often?

The second half of 2007 saw violent moves in credit spreads. In the fallout, there has been much discussion about how to estimate the probabilities of these severe events, but few conclusions have been obtained beyond the fact that historical data is often unhelpful. In this article, Richard Martin argues that the credit default swap curve prices much of this information in, and that these probabilities can be extracted using a Levy-based structural model without recourse to historical data; by contrast, reduced-form approaches cannot do this. He also introduces the concept of 'market-implied value-at-risk'

A central question in managing the mark-to-market risk of a credit trading book is ascribing probabilities to big spread moves or, equivalently, finding mark-to-market value-at-risk at high confidence levels. It is worth winding the clock back to early July 2007 (fortunately one has such luxuries when writing articles) to get an idea of what different models say, or said, about such events.

Suppose, for added realism, that you are the quant responsible for managing desk risk, and the head trader