Veron emphasizes that it may be most useful to require those issuing securities, and asking to have them rated, “to disclose more standardized and audited information about their risk factors and financial exposures.” This new regulatory rule would apply to ratings for corporate bond issues, for financial securities being rated, and also for sovereign debt being rated. Credit rating agencies could be required to conduct their ratings with whatever methods they choose–but based only on the publicly available information.
I think that we need to consider more deeply what these ratings measure, and how it relates to risk.
It helps to think of bond investors as writing an out-of-the-money option to default. This is the way we thought of mortgages at Freddie Mac. If you put down $20,000 on a $100,000 house, then your option to default is “in the money” if the price of the house falls below $80,000. When the price is $100,000, your option is out of the money by $20,000.
I think of a AAA rating as a statement that the option is very far out of the money. A B- rating says that the option is much closer to being in the money.
As an aside, there are other dimensions to risk. Sticking with the example of mortgages, an $80,000 mortgage on a condo tends to be riskier than an $80,000 mortgage on a detached home. That is because condos are “high-beta” properties, which tend to rise more in bull markets and fall more in bear markets. Another risk factor is non-owner-occupancy. Someone who invests in a property solely to earn income will tend to exercise the default option more ruthlessly than will a family who bought the house because that is where they wish to live.
Suppose that a rating for a security is a measures how far the default option is out of the money. Buyers of securities can make whatever use they like of this measure. But does it help regulators?
I can think of two reasons that the rating might not help. One is that a regulator, such as the FDIC, is in a position of having written an out of the money option on the institution. The institutional option may be closer to being in the money if the institution holds one AAA security than if it holds a diversified portfolio of B-rated securities. In fact, this is what led to a lot of regulatory arbitrage under Basel I and to the Basel II concept of institution-wide model-based measures of risk.
An even larger issue is that the risk of an option is not summarized by how far the option is out of the money. The risk comes from the sensitivity of the option to underlying variables. It is that sensitivity to underlying variables that the regulators ought to be concerned about. For that reason, I do not see any easy way to reform the credit rating process to make credit ratings a sufficient statistic for regulators.
READER COMMENTS
Jon Leonard
Nov 29 2011 at 12:46pm
That’s certainly a part of the valuation, but not all of it. Individual credit scores don’t take specific (proposed) loan details into account, but they still have significant predictive power.
Some of that is because options are independent transactions: If I exercise an exchange-traded option, that doesn’t at all impair my ability to buy more in the future. In contrast, someone with a bond or loan default in their history will have changed their ability to get loans (or at least changed the interest rate). Alternately, the cost of defaulting on a loan (exercising the option) includes reduced access to future funding, which makes exercise less likely than in a standard option scenario.
This doesn’t make the analysis any simpler, of course.
The Snob
Nov 29 2011 at 1:52pm
This feels just like the attempts to re-inflate the housing bubble. The ratings agencies failed as spectacularly as a certain famous ocean liner, and here we are arguing about what color paper to print the reports on.
The incentive structure here is all wrong. First, security sellers shouldn’t be the ones choosing the rating agency. That should be assigned randomly or better yet chosen by prospective buyers. Second, you can’t outsource risk assessment to a third party with no skin in the game and expect good results. The agencies need to have something bigger at risk than their pride. Perhaps require them to take a position in the securities they rate in proportion to the rating they give.
R Richard Schweitzer
Nov 30 2011 at 4:33pm
Much as this would disappoint v. Mises, risk involves probabilities, two kinds actually; the likelihood of occurence and the possibility of non-occurrence (positive and negative probability)
The algorith for each must be separately derived, because the factors that might lead to occurence vary from factors that might prevent occurence.
If memory serves, Turing noted this might have particular application to finance (I don’t recall that he framed it in risk terms).
Comments are closed.