Credit Ratings in the Chilean Fixed Income Market

Credit Ratings in the Chilean
Fixed Income Market:
Some Empirical Observations
A. Cifuentes, V. Charlin and E. Bone
Centro de Regulación y Estabilidad Macrofinanciera, CREM
Faculty of Economics and Business
University of CHILE
Santiago, CHILE
January 2014
Background
Fixed income markets are highly regulated
•
What institutional investors can and cannot buy is dictated by the
credit ratings
•
A bond issuer has to obtain two ratings before it can market a
security
•
It has become widely accepted --and demonstrated-- that the
rating agencies played a significant role in the subprime crisis
***
Is it reasonable to have a regulatory
framework based on ratings?
Questions ?
Are ratings (as they are defined now)
proper metrics to assess credit risk?
Should the rating agencies that played a
role in the crisis be allowed to operate in
Chile?
The goal of this investigation is to shed some light
into an aspect that so far has received little attention
We seek to explore whether the ratings given by the
three leading rating agencies in the Chilean market
are indeed different
A casual observer would probably notice a high degree of
"agreement" in terms of the ratings
This issue is relevant for at least two reasons
•
If two ratings are required, but they are always the
same, does the second rating offer an
investor/regulator any additional value?
•
Given the fact that many subjective elements come into
play when issuing a rating--not to mention that the
agencies proclaim to have different methods and
benchmarks--one often wonders: should there be a high
level of agreement?
This market is dominated by three rating agencies
(Moody's, Fitch, and Standard and Poor's)
The ratings reflect credit risk (ability and willingness to
pay); they do not capture market risk, liquidity risk, or
the possibility of fraud
Standard and Poor's (S&P) and Fitch base their ratings on
a "probability of default" concept
Moody's claims to give ratings based on an "expected
loss" concept
If P is the default probability, then the expected loss
(EL) associated with such default is
EL= P(1-α)
where α denotes the recovery rate
Suppose I buy a bond that will pay $ 100 in a year
Let us assume that p (probability of default) is high, say, 90%
[1] The S&P rating should be “low”
[2] But the Moody’s rating (based on the expected loss concept, EL)
could be “anything” depending on the recovery rate
EL = p (1 – α)
If α = 10% then, the EL = 0.9 x 0.9 = 81%
If α = 99.9% then, the EL = 0.9 x 0.001 = 0.09%
Based on this: Should we expect a high
degree of agreement in the ratings?
Ratings Categories and Symbols
Default
Probability
Expected
Loss
Can we say that AAA = Aaa?
Or that
A+ = A1?
Or that BBB+ = Baa1 ?
Data and
Findings
Exhibit 4. Values of kappa and its corresponding 95%-confidence
intervals for both structured products and corporates.
Exhibit 5. Values of weighted kappa and its corresponding
95%-confidence intervals for both structured products and
corporates.
Some Unsettling
Regulatory
Issues
The results are clear: the level of agreement for both,
corporates and structured products, is extraordinary
One possible explanation is that predicting the capacity of a
company to pay its debts is a remarkably objective endeavor
Another alternative is that the agencies after years of
competing with one another have learned to "calibrate" their
methods not to outdo each other
Whatever the explanation, we prefer to leave up to the
reader the task of deriving his/her own conclusions
One thing, however, is obvious: the benefits of having
two agencies analyzing each transaction when we know
that the outcome will be, most of the time, a close
agreement, seems dubious
Finally, a deeper reflection:
• The findings of this effort are just one additional
piece of evidence--in a long list of exhibits--to
think about reframing completely and from
scratch the entire credit ratings concept
• Whether the findings of this study are a
peculiarity of the Chilean market, or they reflect
an overall global trend, is something we intend to
address in the near future
• Next project: Mexico & Brazil