Why credit default swaps don’t stand alone as a predictor of failure
xpanding on a recent blog post at
www.kamakuraco.
com, which in turn
was an update on
a 2007 article in
Risk, we are delving into one of the
most important insights to come out of
the credit crisis: How the probability of
government bailouts causes the potential losses of different liability holders to
diverge sharply. A similar kind of division can be seen in the trends of default
probabilities and credit default swaps.
We compared five-year composite
senior debt CDS quotes for Citigroup
and Fannie Mae (FNMA), as reported
by Markit Group, with reduced-form
E
declaration of a D or SD rating (if the
latter becomes permanent); or a credit
event under the International Swaps
and Derivatives Association CDS language.
The five-year default probability is
assembled from 60 different logistic regressions that predict both the spot one-month default probability and 59 forward one-month default probabilities.
The five-year probability is quoted on
an annualized basis comparable to the
CDS quote.
For FNMA, on September 5, 2008,
the KRIS probabilities were consistent with a 42.92% five-year cumulative probability of default. Annualized
default probabilities on that day were
in excess of 21% for the one-year and
insured debt. In the case of a bailout,
these rates would be near unity. Hence,
CDS rates would be significantly below
the probability of failure. Equity holders’ values do not depend on this recovery rate.
A similar pattern can be seen in
Citigroup. From January 1, 2007 to
March 31, 2009, the five-year default
probabilities peaked above 25%. The
senior debt CDS quotes topped out at
6.35% on March 31, 2009. Again, the
CDS spreads obscured the probability
of failure, because they also included
the probability, contingent on effective failure, of senior debt holders being rescued. Common shareholders, by
contrast, care about the raw probability
of failure because they’re unlikely to
Default Probabilities and CDS Quotes
default probabilities like the CDS/
default probability ratios available on
Thomson Reuters’ 3000Xtra service.
The CDS quotes are on a spread basis
for senior debt with “modified restructuring” documentation. The default
probabilities are the five-year Jarrow-Chava version 4. 1 reduced-form default probabilities from Kamakura Risk
Information Services (KRIS). These
predict the probability of failure from a
combination of financial ratios, macroeconomic factors, equity market returns
and company size.
These models are benchmarked on a
“failure flag” that is the earliest to occur
of the following events: Chapter 7 or
11 bankruptcy; de-listing for reasons of
obvious financial distress, such as failing
to pay fees or file financial statements;
10.61% for the five-year default probability. The five-year CDS quote, by
contrast, was only 38 basis points.
Which was the best estimate of the
failure of FNMA, broadly defined? From
common shareholders’ perspective, the
default probabilities were better predictors of their almost total loss (although
the stock continued to trade at slightly
above $1). The 38-basis-point credit default swap represented something completely different – a combination of the
probability of failure, risk aversion and
other factors, and the probability of the
rescue of the senior debt holders, who
were bailed out. That’s why the two
time series can and do diverge.
Another way of understanding this
difference is that CDS rates also depend
on the actual rate of recovery on the
be bailed out. That’s why default probabilities are much more important than
CDS spreads to investors in common
stock, preferred stock and even subordinated debt.
Best-practice risk management requires transparency with respect to both
the probability of failure and the potential recovery for each class of liabilities.
Both CDS quotes and modern default
probabilities are essential for best practice risk management.
Robert Jarrow is professor of investment management
at Cornell University’s S.C. Johnson Graduate School
of Management and managing director and director
of research of Honolulu-based risk solutions company Kamakura Corp. Donald van Deventer is founder,
chairman and CEO of Kamakura. Their last contribution to Risk Professional was “VaR Versus Put
Option” in August.
