Bank Stress Test (1) or default predictors

"You have been retained by the US Government to figure out some key issues related to
the current financial crisis."

This is how the final exam at Booth School of Business might start. In fact it just did, in Pietro Veronesi's "Mathematical Models of ..." class.

Here are some of the interesting points from this exam that I've found, which are closely related to US banks "stress-testing" that US regulators are about to disclose.

On the first graph default probabilities around two major recent default events are shown. This is based on historic time-series of spreads on credit-default swaps (CDS) and it covers the period from January 2007 till December 2008.



You can see that at the moment of Bear Stearns default the probability to default in one year spiked at around 12-19% level. At the moment of default both Bear Stearns and Lehman Brothers were priced as if both are exceptionally close to default. While Bear Stearns CDS spread afterwards went down (probably because of obligations transfer) the spread on Lehman Brothers still suggested very high probability of default in one year. Just before the moment of actual default the implied probability to default in one year reached about 17%.

We can’t say definitely based on CDS market signals whether risk aversion is up for marginal (price setting) market participants or “true” probability of default indeed went up for some fundamental reason. However, the graphs dynamics is illustrative showing that CDS spreads do allow to obtain valuable information about the state of the financial company not reflected on other securities markets.

My hypothesis is that "better than the rest of the market" prediction power of CDS is due to the risk management at major financial institutions who have to purchase protection on the company debt they hold. Since those players are so big and well-connected they have an advanced knowledge about “true” state of the companies way before public information is available. Once they see that the condition of company is about to worsen they start buying more protection since selling bonds at those times could be very hard due to the lack of liquidity on those positions. This increase in demand for protection causes the spread to widen.

It also triggers chain effect since the banks who sold CDS protection earlier for low price now are losing money and they have to hedge their prior exposure by buying new CDS contracts and so on.

There is also a "speculation" hypothesis flying around these days, and it might be somewhat true since markets are still liquidity constrained. However, I still think it is mostly demand driven rather than supply constrained.




On the next graph there are time-series for derived cumulative default probabilities for Citigroup and Wells Fargo.
It covers the period from January, 2007 till March, 2009. The probability of default for Wells Fargo around Lehman Brothers bankruptcy was higher than the one for Citigroup, recently it tends to be the other way around. The merger of Wells Fargo with Wachovia seems to have positive effect.

This is of course risk-neutral probability, which sets the upper bound for the possible range of risk-natural probability. Still, this is extremely high and it looks similar enough to the case case with the bankruptcies of Lehman and Bear Stearns

You can also clearly see that Lehman Brothers collapse triggered a huge spike in CDS implied probability of default and there were no market signals at all BEFORE Lehman Brothers collapse. Even CDS spreads didn't show that financial Armageddon is about to be launched.

The problem with Citigroup becomes even clearer if we look at term structure of marginal default probabilities (not shown here). The term structure for Citigroup is strongly inverted, much more so than in the case of Wells Fargo. For Wells Fargo the slope is almost linear, although it is also slightly inverted. This again strongly reminds of Bear/Lehman case when such invesrion surfaces right before BSC collapsed.

I will look into a different way to evaluate the health of financial institutions in the follow-up post.