Credit Risk Management  

   Last revised: March 02, 2002

New, this month:

 7/28/00 Ask Dr. Risk!: "Managing Credit Risk on a Long Term Lease" and "Predicting the
Probability of Default."
 7/28/00 Dr. Risk's Derivatives DictionaryTM: EDFTM, Expected Default FrequencyTM, KMV.
 7/28/00 Dr. Risk's Links

On this page:
Ask Dr. Risk! | Books | Links | Presentations | Derivatives DictionaryTM

Ask Dr. Risk!

Dr. Risk promises you at least a brief response to your important question, as soon as he has a free
moment. A question of sufficiently general interest to make it into the 'Zine, tends to generate a more
comprehensive response. All questions and answers become the property of The William Margrabe
Group, Inc. 7/28/00 

Managing Credit Risk on a Long Term Lease  (7/28/00)

Dear Dr. Risk – I am working as treasurer/risk manager and have the following problem. My company
intends to buy an dedicated asset and operate it for 15 years and supply the products to our Client under
a long term (15 yr) contract, wherby Client is paying for the capital costs of the investment plus the
variable cost and marging for the products. This client however has a rating of BB, which is quite low. Can
you provide me with ideas on products and pricing techniques to hedge the counterparty risk for this
casus. – Jan

Dear Jan – Your specific situation depends on specifics about which Dr. Risk has no information, but here
are some general comments about your sort of situation. 

It sounds as though:
(a) You have made essentially a secured loan to your client (capital costs), probably fully amortizing, with
a level payment, much like a residential mortgage loan. 
(b) You've bought an unsecured annuity from your client (margin).
(c) The variable cost is irrelevant, because if the client defaults, he's bankrupt and you can stop
production, avoiding the variable cost. 
(d) You bear some risk related to output quantity and margin.
Initially, let's ignore the equipment as security, as well as the quantity risk. We'll touch on them, later. 
Hedging might conceivably be the prudent way to go to manage risk associated with the loan and the
margin. At first glance, the actual  loan payments and annuity are the sort of cash flows that you could
replicate with the client's debt, and the promised loan payments and annuity are flows you could replicate
with AAA debt. If so, then your ability to construct a hedge depends on your availability of hedge
instruments. 

However, bankruptcy law handles lease payments, interest payments, and principal differently, and that
may complicate the replication problem beyond solution. Even under the best of circumstances,
appropriate instruments may not be available, and constructing the hedge is not a trivial matter. 

The equipment as security for the capital cost payments complicates the formation of the replicating
portfolio. It seems clear that you will not find a perfect substitute for it in the marketplace. 

Dr. Risk cannot think of any way to hedge the quantity risk from the profit margin, without simply selling it
in the market.  

You might be able to buy credit protection from a credit derivatives dealer. It would face precisely the
same problems that you would face, but is likely to have greater expertise and market access, because it
has been in the business. 

You could more easily buy insurance or a guarantee from an insurance company or a credit guarantor.
Pricing is simple, but obscure. You describe the coverage you want, and prospective vendors name the
premium. It’s not clear which actuarial model they would use, so it’s hard to duplicate the model. They
diversify away the risk, so you probably couldn’t duplicate the protection by bypassing the insurer or
guarantor.

If this protection is valuable, it doesn't come cheap. At some point the cure would be worse than the
disease. Consequently, many suppliers in your situation just play with fire, and many get burnt. 

Please let us know if you need names of insurers, guarantors, or credit derivatives dealers. – Dr. Risk

7/28/00  Predicting the Probability of Default  (7/28/00)

Dear Dr. Risk  – Apart from KMV, what other models are being used  for forecasting the probability of
default. Could you please throw some light on the relative performance of these models. Do you think that
the credit rating agencies are doing a good job in terms of prediction of bankruptcy risk? How would you
rate the rating agencies in this regard? – Amiyatosh

Dear Amiyatosh –  Dr. Risk can think of three main methods to forecast default, or compute or estimate
the probability of default:
1. KMV / Geske
2. Strippring risk neutral probability of default from credit spread
3. historical relative frequency of credit rating migration 

Method 1. 
KMV has a way of taking accounting and market data about one company, producing the inputs that the
Black-Scholes-Merton (1973) model requires, and then outputting the implied probability of default, N(-
d2). Bob Geske has extended this to the simplest term structure of probabilities of default by using the
Geske (1979) compound option pricing model. Using essentially the same data that KMV does, but
refining things a bit, he produces a probability of default in the nearer period, then a conditional probability
of default in the next period given no default in the first period.
Method 2. 
The second method is stripping the risk neutral probabilities of default from the price and coupon
information on one name's debt issues of different maturities. Thus, given a one-year corporate bond and
one-year sovereign debt in the same currency, one can make a few leaps of faith and deduce the risk-
neutral probability of default during year one. Then you add the two-year corporate and sovereign to the
mix and deduce the company's conditional probability of default during year two. And so on, and so on,
and scooby, dooby, doo.

Method 3. 
The third main way to go is to use historical data for different credit ratings. Moody's and S&P produce
and publish credit rating transition matrices. They describe the historical relative frequency of
occurrences, such as "in one year  a BB bond became a AA bond 0.02% of the time." They also describe
things, such as "in one year a BB bond defaulted 0.5% of the time". Of course, Dr. Risk just made up the
specific numbers. If you want the real ones, get the tables from the agency.

How well do these methods perform? Dr. Risk saw Peter Crosbie's presentation on credit risk at a
conference, recently. He had graphs that indicated that methods 1 and 2, using market data, showed
marked deterioration in implied probability of default in advance of actual default. Dr. Risk doesn't have
the impression that either of those methods was the undisputed king of early warning systems, and each
has its methodological weaknesses. However, clearly, it appeared that both methods provided leading
indicators of changes in credit ratings, and that is consistent with Dr. Risk's priors. – Dr. Risk   
P.S. What does your expert colleague at Cornell and my friend, Bob Jarrow, have to say about all this?

Modeling the credit risk of credit derivatives  (6/28/00)

Dear Dr. Risk – How do practioners determine the amount of the exposure offset created by a long
default swap, where the maturity of the default swap is less than the reference/underlying asset? I now for
regulatory capital purposes there are policies set forth (eg: FSA in UK)  which outline the conditions but
was interested to know want how bank's actually record the offset? – Peter

Dear Peter – In my experience, a default swap always matures when the underlying/reference note/bond
matures. A total return swap may mature before the reference instrument matures.

You asked specifically about the impact of the default swap on exposure. A typical measure of exposure
is replacement cost, or market value. I'd say it's probably still more common to look at exposure, deal by
deal. 

In any case, a logical way to incorporate a short-dated total return swap on a long-dated note or bond into
computation of credit exposure is to computer credit exposure, both before and after you add it to the
portfolio of deals with the relevant counterparty. The change in exposure is simply exposure "after", minus
exposure "before".  

It may make more sense to look at things in the proper portfolio setting, not deal by deal. Perhaps,
consider the portfolio of all deals that should net, upon default. Also, exposure isn't the only measure of
risk that one might use. Credit VaR comes to mind, although that has its drawbacks. – Dr. Risk

Pricing Credit Risk  (6/28/00)

Dear Dr. Risk  – We are looking for a good book on Pricing for Credit Risk and details of how and where
to obtain it. – Sarie

Dear Sarie  – Dr. Risk has links to books on credit derivatives

at [Link]
Also, the Jarrow-Turnbull theory book that you'll find there covers pricing of credit derivatives at a
sophisticated level. – Dr. Risk

Expected Loss on Retail Loans  (6/28/00)

Dear Dr. Risk – I am interested in learn more about the probability of default, de loss given default rate,
the expected loss in credit risk, particular for retail loans. Where can i find or download information about
this topics? – Pablo

Dear Pablo  – You could do a lot worse than looking at the resources available
at [Link] Dr.
Risk needs a more specific question to provide a more specific answer. For example, Dr. Risk doesn’t
know what you mean by retail loans. You could mean real estate mortgage loans, commercial credit, etc.
It’s not clear whether you are interested in theory or empirical work. If it’s theory, you can pick from
several possible models that encompass elements of your question, mainly reduced form models, such
as Jarrow-Turnbull and Duffie-Singleton. If it’s empirical work, the main data sources are the credit rating
agencies. For individual credit, you might want to see what Fair Isaac has to say. – Dr. Risk

7/28/00  [more] (7/28/00)
Dear Dr. Risk – I am intersted in comercial loans default and consumer credit too and i want to find some
information about how can i model the probability of default of a portfolio form by consumer credit, and
comercial loans. The way of how can i provisioning, working with two eyers historic information about the
way of modeled loans default by internal rating categories trougth credit scoring. Where can i find
information about those reduce form models you mention such as Jarrow- Turnbull and Duffie -
Singleton ? – Pablo

Dear Pablo  – Moody's has a mid-market commercial loan model. You can probably find some information
about it their web site, with a link from the Credit Risk page [Link]/[Link]. Dr.
Risk is not sure what they give away.

Loan Pricing Corporation is in this business, too and has a mid-market product. See what they have
at [Link]

Dr. Risk has already mentioned Fair, Isaac, which is the main source for information about credit scoring.
It has a link from [Link]/[Link].

Dr. Risk can't refer you to a public source on modeling the probability of default of a portfolio of consumer
and commercial loans. He's not sure what you mean. If you have a portfolio of loans, one may default and
the rest may not. Does that mean the portfolio defaulted?

You can find information about the Jarrow-Turnbull model in their book, which appears
on [Link] You might look at some of Duffie's downloadable
papers at [Link] a sufficient description of the
Duffie-Singleton model. – Dr. Risk

Where did the CreditRisk+ Technical Document go?  (5/28/00)

Dear Dr. Risk  – I am a final year Finance/Maths student who is currently doing a research

As such, I would like to ask you a few questions: Firstly, would you know where I could get a copy of the
CreditRisk+ Technical Document? It appears that it has been taken off the CSFP website.

I was also hoping to get your opinion on the current models and where you think improvements should be
made and where the focus should lie. – Kristin

Dear Kristin – Dr. Risk checked the link from The Derivatives 'Zine to the CreditRisk+ technical document
and confirmed that it was broken. Thanks for informing us. Next, Dr. Risk contacted Tom Wilde at CSFB,
because he was the prime mover who innovated CreditRisk+. (GARP recently honored Tom as "Risk
Manager of the Year", largely for that service to the risk management community.) Tom graciously
informed us of the new location, [Link]

Concerning Dr. Risk's opinion on current models: Dr. Risk wouldn't want to deprive you of the opportunity
to dig into the models and develop your own opinions. However, at the risk of coloring those opinions,
here are some questions that Dr. Risk would try to answer while modeling credit risk. The answers to
these questions will help shape a useful model:

1. Whose risk are we trying to measure? Shareholders'? Directors'? Traders'? Society's?

2. What do these people want to know about credit risk?
3. Does it make sense to differentiate credit and market risk? Does it make sense to combine them?
4. What should those people do about the risk?
5. Does the model provide any hypothetically useful answers to significant risk management
questions?
6. What is the tradeoff between risk and reward, and does the model address it?
7. How good are the required data? How valuable is the model, given the available data?

– Dr. Risk

P.S. I wasn't able to open the documents that I downloaded. My old versions worked fine. Let me know if
you have a similar problem.

How has the level of whatchamacallit been related to you-know in the

past?  (5/28/00)

Dear Dr. Risk –  what is the historic correlation pls between interest rates and credit, assuming underlying
credit is investemnt grade. – roger
Dear roger  – Dr. Risk is many things, but – alas – he's not a mindreader. Could you please provide a little
context and explain why you want to know what you're asking? That way, Dr. Risk won't waste your time
with detailed answers to questions that don't interest you. 

Let's take your question, almost word by word. 

 What do you mean by historic? Which time period? What part of the credit cycle?
 What interest rate? Interest rates in general? The yield on investment grade issues?
 What is "credit"? Credit spread? Credit quality? Credit quantity? 

Please help Dr. Risk help you. What are you trying to do? Why do you want this historical correlation?
What model are you going to put it in? A credit derivatives pricing model? A credit risk management
model? – Dr. Risk

Is the M&M theory of the cost of capital all wet?  (4/28/00)

Dear Dr. Risk  – I own shares of Kelda Group, which owns Yorkshire Water Services. It's definitely not
"New Economy", and I'm pessimistic about its long-term prospects. Recently, I heard that Moody's is
probably going to downgrade its bonds from A2, so its yield spreads widened. An analyst at Moody's said,
"At the moment debt is cheaper than equity for water companies ..." Kelda is thinking about securitising its
revenue stream to raise money.  (Aline van Duyn, "Credit reratings may hit water company
bonds," [Link], 4/12/00) Do you agree that that's best? – Roy Poseidon 

Dear Roy – When you make a value judgment, such as what's "best", you have to first pick a set of
values. In this case, stakeholders include 

 shareholders
 original bondholders
 new bondholders
 the bankruptcy bar
 Inland Revenue

Dr. Risk presumes that you would prefer to use your values, so your question is, "Do you agree that that's
best for me?" or "... for original shareholders?" However, we'll have to consider the impact on other
stakeholders to deduce the impact on you. 

Your question raises three of the major issues in corporate finance: 

 optimal investment 
 optimal capital structure
 optimal dividend policy

Of course, Dr. Risk will look at these from your point of view, not from some ivory tower. 

In any event, the buyers of the new bonds backed by the revenue stream should be indifferent to this
scheme. If Kelda sells the bonds properly, these bondholders should pay a competitive price for their
bonds. Good security, hence a high price. 
The bankruptcy bar benefits from any plan that significantly raises the probability of Kelda's financial
failure, landing it in bankruptcy court. Roughly speaking, a decrease in a corporation's equity as a
percentage of assets makes bankruptcy more likely. 

The Inland Revenue (the U.K.'s version of the U.S.'s IRS) is always sitting at the table, pounding the butt
ends of its knife and fork against the table, demanding its piece of the pie that someone else baked. Dr.
Risk doesn't know much about taxation in the U.K., hence would not care to speculate on the impact of
Kelda's financing on the collections by the boys and girls at Somerset House (Inland Revenue's HQ).
However, in the U.S., corporate finance decisions can significantly affect the corporate and personal tax
bite out of any business operation.

You're not clear on Kelda's investment and dividend plans, hence on the destination of the funds raised. If
Kelda is such an "Old Economy" investment, perhaps this is part of a scheme to disinvest and pay
dividends or buy back shares. Either way, the same collateral would secure even more promises to pay,
and the old bondholders would find the new bondholders in line ahead of them when it came time to seize
collateral to get paid. Clearly, this would be bad for existing bondholders, neutral for the new bondholders,
and good for shareholders, such as you.

Perhaps, Kelda plans to invest the funds in new plant and equipment, including prudent repairs on the
old. Again, the new bondholders would be indifferent. The old bondholders would lose some collateral,
but would gain a claim on new assets. Dr. Risk can't be sure of the overall impact on the shareholders
and the old bondholders. The key issue is the probability distribution of asset value at the time the debt
matures. If asset value exceeds debt, then the corporation will pay the bondholders in full. Otherwise, the
corporation may default. A purchase of additional assets should tend to shift the probability distribution of
future asset value to the right, which would be good for the bondholders. However, the investment might
also affect the variability, which could be good or bad for the bondholders. Increased variability works
against the bondholders, because it increases the value of the shareholders' (put) option to default.
Increased variability could result from investment in a highly risky asset. Investment in an asset with
negative correlation with the original assets decreases the variability of asset value. If that risky, future
asset value will likely exceed (fall short of) the debt's face value, then decreased variability makes the
bonds more (less) valuable, and increased variability makes the bonds less (more) valuable. – Dr. Risk

Waiter, would you recommend the Wiener ... or the Poisson?  (4/28/00)

Dear Dr. Risk  – I am familiar with the properties of Wiener Processes, but I would like to have a good
reference (either book or article) explaining the basics of Poisson Processes. I need to do some Monte
Carlo modelling of Jump-Diffusion processes, and I want to make sure I get the step size right to ensure I
only get one jump (at most) per period. – Dr. B.

Dear Dr. B. – Dr. Risk has three suggestions, all on Dr. Risk's Bookshelf and through [Link]: 

1. They say nobody ever forgets his first time. For Dr. Risk, the first time he looked at Poisson's
Theorem, the Poisson distribution, and Poisson processes was in Boris Gnedenko, The Theory
of Probability (B.D. Seckler's translation of the fourth Russian edition. I found his discussions
sometimes concrete, sometimes abstract, and always clear. You can now find Igor Ushakov's
translation of the sixth Russian edition at [Link] .
2. Howard M. Taylor and Samuel Karlin,  An Introduction to Stochastic Modeling.  The explanations
are also clear and thorough, and it's about stochastic modeling, which seems to be what you
want.
3. A more theoretical book is Reza Iranpour and Paul Chacon,  Basic Stochastic Processes; The
Mark Kac Lectures. The book is based on the lectures at USC in 1982-1983 of Kac, one of the
 greats in probability theory. Sadly, this title is out of print, but Amazon will search used bookstores
for a copy.

Dr. Risk hopes one of those books is what you want. – Dr. Risk

Risk vs. Reward in the "Full Monte" for Credit Metrics  (10/28/98)

Dear Dr. Risk  – i have [a] question ... concerned to risk-return-analysis Situation: If i use CreditMetrics
(full mark-to-market version), the expected return of one exposure can be calculated as

 [(mean value - current value)/current value] - cost of funding

 Afterwards i can compare this return with the risk contribution of the exposure to the portfolio. The result
is a relative ratio

My question: If i use CreditMetrics in a reduced form, that means i only consider default or non default
(like CR+), i get a portfolio distribution of (potential) losses. I can analogous calculate the risk contribution
for each exposure. What a measure for the expected return of each exposure should be chosen? – Mike

Dear Mike – In all seriousness, Dr. Risk's answer depends on his best guess of what you're really trying
to accomplish, which you didn't state and which isn't obvious to Dr. Risk. Dr. Risk guesses that you want
to develop a methodology  for making credit decisions by a sort of marginal analysis. You want to accept
the loans that have a high ratio of reward to risk, reject the loans that have a low ratio of reward to risk,
and find the dividing line between the acceptable and unacceptable loans. In order to do this, you clearly
need a measure of reward and a measure of risk. You've adopted the marginal contribution to credit
exposure, as CreditMetrics computes it, as the measure of risk. For CreditMetrics "Full Monte" Carlo
simulation of mark-to-market, you're willing to use the excess of an asset's expected rate of return over
cost of funding as a measure of reward. You want to know what measure to use when you are
considering only losses. 

Dr. Risk could easily and truthfully answer, "The same measure of expected excess return over cost of
funding will be just as valid in the 'reduced form' case as it was in the 'Full Monte' case." Unfortunately,
that would be both true and misleading. The basic approach you want to take has the important and
sometimes over-riding advantage of producing a decision rule, but the disadvantage of being "ad hoc"
and even questionable. In academic terms, it doesn't have a firm theoretical foundation. In layman's
terms, it doesn't really make sense. 

I see a basic similarity between what you want to do and what Harry Markowitz did  with portfolio theory –
you want to pick the optimal spot on the tradeoff between reward and risk. So you're off to a good start.
Your choice for reward is the same as Markowitz's. 

Your problems start when you pick a measure of risk (VaR) that Dr. Risk cannot imagine figuring directly
into the decision maker's utility function. Markowitz chose variance as his measure of risk and Tobin
justified it two ways, each sufficient by itself: (1) if the asset price distributions are normal, then variance is
the natural and only measure of risk, (2) if the investor has a quadratic utility function, then you needn't go
beyond variance to analyze risk. VaR as a rational measure of risk? Questionable, at best. (If you can
come up with some sort of reasonable and rational argument for that, please call Dr. Risk at any hour of
the day or night. This will be momentous news.) 
If your measure of risk is questionable, then your results will be questionable, regardless of your measure
of expected return. I would suggest going back to first principles, figuring out how your decision maker
makes decisions (e.g., expectation of quadratic utility of wealth) and/or how does the world  work (e.g.,
normal distribution of returns) and deduce your measures of reward and risk from there. Please, next
time, bring Dr. Risk into the problem closer to the start, rather than at the end of seems to be a blind alley.
– Dr. Risk

How safe are dollar deposits in various banks around the world?  (9/28/98)

Dear Dr. Risk  – I would like to know wich are the risk of the dollar deposits in:

 1- A International [i.e., foreign] Banking Facilty in the US

 2- In a us bank branch in London
 3- In a us bank in Hong Kong
 4- In a Mexican Bank in Mexico

– Rene

Dear Tommaso – While I'm asking some experts on this topic, please answer me this: Why do you care?
Idle curiosity? – Dr. Risk

Dear Dr. Risk – I work for the Central Bank of my country, which have recently been upgraded in his
credit rating and is seeking markets to invest. Also is a good subject to give in my class of International
Banking and a good example for my students in the usage of the internet as a research tool. – Tommaso

Dear Rene – You closed off discussion of the thorny issue of devaluation by denominating all the
accounts in dollars. Good move.

I spoke with a Wall Street professional ("Señor Wall Street") and a source closer than I to Alan
Greenspan ("Deep Vault") The key issues that came up were (a) the sanctity of property rights, (b) the
banking culture , (c) capital requirements, and (d) other regulations. Obviously, an "International Banking
Facility" might be a bank of the U.S., U.K., Japan, etc. (you get the picture), or Mexico, Nigeria, etc. (you
get the picture). So let's divide that into 

1a - A British Banking Facility in the U.S. 

1b - A Mexican Banking Facility in the U.S. 

All other things being equal, I would say that you put 2, 3, and 4 in the right order. The British Banking
Facility in the U.S. would be right up there with the U.S. bank branch in London, maybe a little ahead or
behind. I'm not sophisticated enough to know for sure which is better, and maybe nobody is. The point is
that U.S. and U.K. banks are – with all their imperfections – relatively reliable wonders of the modern
world, and we're fortunate to have them. Hallelujah! (Senior Wall Street thinks I'm too sanguine about
prospects for U.S. banks during a recession with loan defaults and praises Swiss bank accounts, backed
with gold reserves.) 

The Mexican Banking Facility in the U.S. would go between the U.S. bank in Hong Kong and the Mexican
Bank in Mexico. Putting money in a Mexican bank has been a little like drinking Mexican water. Mexican
banks in Mexico have a long record of ... uh, not being Swiss banks, and there is only so much that U.S.
regulation can do to counter that. Think of U.S. regulation of a Mexican bank's U.S. branch as akin to an
iodine tablet dissolving in a canteen of Mexican water. In theory, the iodine should purify the water, but
would you really drink it, if you could drink New York City tap water, instead? Would you really put your
money in a Mexican bank in New York, if you could put it in a New York city branch of a U.S. bank? The
Hong Kong branch of CitiBank isn't quite as secure as a New York branch. However, the Communist
Chinese have not yet proven that they are going to loot Hong Kong, and there is reason to hope that they
may never do it: people aren't papering their walls with Hong Kong stock certificates. So, I'll give the U.S.
bank in Hong Kong the benefit of some doubt. 

Here are comments from the experts: 

I. Señor Wall Street is a Mexican national who works on Wall Street. He rated the credit quality, as
follows:

1. (best) US bank in US
2. Mexican bank in US
3. US bank in Mexico
4. Mexican bank in Mexico. 

As you can see, I didn't ask him your question, precisely. 

His argument is that US banks follow US practices and have US regulation. The foreign banks may have
some freedom from US regulations and personnel certainly have bad habits from their hometown banking
practices. Any money in Mexico is up for grabs! The US bank in Mexico has to fear some US regulation
and scrutiny, which keeps it in line, and it has some protection, because the Mexican banditos don't want
to anger the US without good reason. The Mexican bank in Mexico appears to have laws and rules to
obey; however, the politically powerful can change them on a moment's notice, to confiscate some or all
of deposits, capital, and equity.

As he explained, "In Mexico, in 1982 they nationalized banks, froze the accounts, and converted the
dollar accounts into pesos at a confiscatory exchange rate. They didn't touch CitiBank, which was then
mainly in the private banking business and had only accounts of prominent Mexicans. Carlos Salinas was
President of Mexico. His brother, Raul, got 10% of every nationalization, hence his nickname, 'Mr. Ten
Percent' ". (For a perspective on the dark side of the Mexican economy, look
at [Link]

II: Deep Vault spoke with me on the condition that I would not identify him or her. I can assure you that
what he says is not an official position of the Federal Reserve Board, any Federal Reserve Bank, or
anybody in the federal government. It's one person's opinion – in my estimation a very bright person's
opinion – and for all I know he wasn't taking his medication. (However, for all I know he wasn't on any
medication.) According to Deep Vault, "As it turns out, the question of supervisory control of foreign
banking entities is a very sensitive issue. Historically (10+ years ago), branches and agencies of foreign
banks were subject to significantly fewer restrictions that US banks, e.g. branching across state lines was
not prohibited.  It is my understanding that US bankers were behind much of the most recent legislation
regarding acceptable/unacceptable behavior by foreign bank branches and agencies.

"Let me now answer the questions addressed to Dr. Risk ... These are quite simple.  I believe the
questions are asking about the riskiness of deposits in particular institutions.  This is a function of where
the capital backing the deposits is located.  If the institution is a US bank, regardless of its location, at a
minimum, it is subject to US regulations (Fed, OCC, FDIC, OTS, and/or state regulators); in particular,
capital requirements.  Additionally, if the branch of the US bank is located outside of the US, it will be
subject to local regs as well.  I cannot speak to the efficacy of these regs.  In some countries, they may be
essentially nonexistent; in others, they may be onerous.  An International Banking Facility is not a legal
entity; it has to do with off-shore banking.  I am not very familiar with these.  Deposits in US banks have
US risk, regardless of where the deposits are located.  That is, the capital backing the deposits is in the
US and is subject to US regulations.  If the deposits are in a Mexican bank, regardless of where the
deposits are (Mexico or the US), these deposits are backed by capital in Mexico.  That capital is not
subject to US regulatory authority.  US regulatory authorities only have authority over branches and
agencies located in the US.

"Does all of that make sense?  In short, authorities in a country only have control over capital located in
that country.  Foreign entities may be allowed to do business in that country, but there is no authority over
the capital (its measurement, adequacy, etc.) that is located in the home country.   For that reason, if you
deposit money in a branch of a foreign bank located in the US, you still bear the risk that the bank may be
inadequately regulated in its home country.  It may go bankrupt and the US will have no control or
authority over that resolution process.

"As to your additional questions ... General Information ... The specific answers to all these queries are in
the Federal Reserve's Regulation K, which details the treatment of US branches of foreign banks and
foreign branches of US banks. Before 1978, there was basically no federal oversight of US branches of
foreign banks.  Each state established its own controls.  There was not coordinated oversight.  The
International Banking Act of 1978 established federal oversight.  In 1991, the Foreign Bank Supervisory
Enhancement Act was also passed. As it stands, a foreign bank must apply to the Federal Reserve.  We
oversee supervision in the US.  The foreign bank may still be chartered as either a state or federal
branch, which will affect which regulator oversees specific functions.

"Historically, the foreign bank could choose to be either insured or uninsured (that is the deposits).  Since
1991, all new insured branches must be capitalized here. When a branch is established, their license will
specify exactly what types of activities it can engage in. Branches can be established as either retail or
wholesale ($100,000 or greater per deposit). Wholesale branches are established in large part to service
high-net worth clients back in the bank's home country, i.e. to offer the safety of dollar-denominated
deposits. There may also be restrictions on the source of deposits.  Some branches may only accept
deposits from non-US residents. Again, all of this is spelled out in great detail in Regulation K.

"Who regulates...It depends on how the foreign branch is chartered and what services they provide.  It
could be as many as five different regulators. There is a chart/matrix that shows which regulator is
responsible for which function in which banks/branches.  It is huge.  I don't know know anyone who has
memorized the entire thing.  It's ugly.

"Deposit structure...It depends on the type of license that the branch has as I indicated above.  Some
branches may be restricted to a single type of deposit.  Others may be able to accept numerous
types.  Again, Reg K.

"London and H.K...Host countries have jurisdiction.  You would have to get information from them about
any specific rules and regs that apply to foreign branches in their countries.

"BIS requirements apply to countries that have adopted them.  In many instances though, specific
requirements are left to the discretion of each country.  Additionally, many countries have different
definitions of capital, i.e. is preferred stock Tier 1 or Tier 2 capital.  I don't want to get off on capital
arbitrage, but I would imagine that certain forms of arbitrage are restricted in some countries and are all
but encouraged in other countries.  This is a topic for another discussion.

"Mexico is an OECD country.  Interestingly, just before the crash in 1994, they were admitted to the
OECD.  Instantaneously, loans to government entities in Mexico had no capital requirements (per the
Basle Accord) as they were deemed riskless.  We all know what happened.  This exemplifies one of the
many problems with the Accord's statutory limits; they are arbitrary.

"Hong Kong...I don't believe it is OECD, especially now that it is back with China.
"Deposits risk in specific institutions...and you want a simple answer...It depends on credit risk and
transfer risk specific to that institution.

"Recommended readings/follow-up if you're really interested...

1. Regulation K (I have been told that it is actually very well-written; i.e., not your typical fed speak. ["Well-
written" is a relative term and highly subjective, as a comparison of  Reg K with the writings of P.J.
O'Rourke shows.]
2. The Fed recently (April 12, 1999) approved the establishment of a US branch of Banco de Credito e
Inversiones S.A. (from Chile).  You might check the Fed's website ([Link]) to see if the
Press Release is there.  I believe it will be.  It briefly explains the process that the Fed went through in
approving the branch license.  Further, it defines the scope of the bank's allowed activities in the US.  In
addition, it states that 'The Board [of Governors] generally also must determine that the foreign bank is
subject to comprehensive supervision or regulation on a consolidated basis by its home country
supervisor.'  Also, 'Chile's risk-based capital standards conform to those established by the Basle Capital
Accord.'
3. Economist article in Jan. 30th issue.  Go to The Economist's website and/or consult your library.  In the
January 30th issue, there is a special pull out section entitled "A Survey of Global Finance," where you
will find a discussion of many of the above issues. Specifically, see p. 8 of the survey for a discussion
about how international investors poured $500 billion in short-term capital into Asian economies with
known deficiencies in capital, financial accountability. See p. 11 about the Federal Reserve and the
foreign branch approval process.
4. Contact Public Affairs at any Fed as they will have some officially prepared responses to many of the
above questions.  I have been told that this is a particularly sensitive area.  We (the Fed) must be very
careful what we say with respect to foreign bank supervision.

"In closing... Well, just imagine what the response would have been if I had truly researched the
topic.  Actually, I found the questions were very interesting.  In fact, they were more interesting that what
ever it was I was working on last week, so they got priority." 

 – Dr. Risk

International Ratings  (7/28/99)

Dear  Dr. Risk – How are derivative instruments rated internationally by rating agencies like S&P? What
are the rating criteria and methodologies for rating these instruments? How are international corporate
bond issues rated by international credit rating agencies? What are the methodologies and criteria used
for rating these instruments? – Roy

Dear Roy – These are good questions, too good for good off-the-cuff answers. The quick answer is that
to the rating agencies obviously have proprietary methodologies, but the methodologies aren't obvious to
outsiders, and they differ both across rating agencies and over time for a given agency. If I look deeper
into this, I'll update your answer. Meanwhile, use my links to the web sites for the rating agencies, such as
Duff & Phelps, Fitch IBCA, Moody's, and S&P. 

Your mention of the word, "internationally", is crucial. International standards for accounting and other
information are generally not as high as those in the U.S. Sometimes, the "standards" could qualify for the
"Derivative Humor" page, except that the jokes are in bad taste. – Dr. Risk
Dissertation Topic: Credit Risk management of commercial loan
portfolio (7/14/99)
Dear  Dr. Risk – I am a financial economics student at the Amsterdam University. Currently I am working
at a thesis about managing credit risk of commercial bank loan portfolio's and I am concentrating on the
application of modern portfolio theory in managing risk and return of loan portfolio's. Although this is a
very exciting subject with lots of different aspects, it is also very complex.

I have read a lot of articles and papers that deal with the issue (including the CreditMetrics and
CreditRisk+ technical documents) and they all provided me with really interesting insights of how credit
risk in commercial loan portfolios could be managed. Yet all authors seem to conclude that there is still a
lot of work and researching that needs to be done in this field. Although this suggests that there are many
opportunities for exciting research, I am somehow finding it difficult to focus on an aspect which would be
value adding to the already existing literature on this subject. Maybe You could give me a few tips about
what would be an interesting aspect to focus on. – Chris

Dear  Chris – Sometimes, the world of credit risk management seems to have more in common with
religion than with science. We have several "approaches" to credit risk measurement, which Anthony
Saunders surveys in his new book, Credit Risk Measurement. You might think of these as analogous to
the world's major religions. In this context, the BIS approach is an state religion, an "established church",
such as England's Anglican Church. It's also something like voodoo, given its lack of sophistication. (In all
fairness, the folks at BIS are not stupid – far from it. They realize the need to enter the 20th century, and
they are looking for a better way to go.)

Few scientists have tried to find testable hypothesis in the approaches, few have tried to compare the
approaches, and no one has provided the evidence to reject any of the approaches. (However, Bob
Geske of UCLA and MKIRisk is doing promising research about the market's ability to predict the
probability of default.) A dissertation that did any of these things would be a great contribution to the
literature. I don't know whether the problem is one of empty theories, missing data, or just the lack of a
recent financial crisis that would allow us to separate good theory from bad.

We don't have good data. The ratings agencies, such as Fitch, Moody's, Standard & Poor's, etc. have
been working in this area for years and collected a great deal of data. However, we still don't have reliable
data on things we would like, such as default rates, recovery rates, or default correlations. Banks have
been making loans for decades, but useful data on credit losses haven't surfaced. Perhaps a bank would
let you analyze its credit data. – Dr. Risk

Pricing credit risk  (1/28/99)

Question: Dear  Dr. Risk – I am a part-time PhD student and my thesis is about interest rate swaps and
credit risk in banking industry. I would be grateful if you could give me some ideas and the current issues
in this area and also it would provide economic value to the banking industry. I have read in fact alot of
the journal articles related to this area, but I find that I got nowhere in the notations and find that some of
the articles are too hard to fully understand. Any advice for me from an experienced Dr. like you? Your
prompt reply would be much appreciated. Warmest regards from Downunder – Vic

Answer: Dear Vic – The intersection of interest rate swaps and credit risk is a busy corner. You're
dissertation should focus on a minuscule portion of the action there, and the same is true of my answer.
Since you mention the tricky notation, I'll guess that you are talking about pricing credit risk for bonds and
swaps. That's my focus.
The literature and industry have three ways of pricing credit risk.

1. Static replication, used to price credit default swaps and total return swaps, is the most useful way for
pricing credit derivatives, but only for some products. Buying protection via a credit default swap is
equivalent to shorting the related floating rate note (FRN) and buying the corresponding AAA FRN. The
related FRN has the same maturity, issuer, credit rating, seniority in bankruptcy, etc. The corresponding
FRN is identical, but has a AAA issuer. The payoff on this portfolio equals a payment stream equal to the
excess of the FRN coupon over the AAA FRN coupon, plus a contingent payment equal to the loss on the
FRN upon default. In equilibrium, the value of the credit default swap should equal the value of the
replicating portfolio.

2. The structural approach to credit default looks at a credit risky bond or swap as a credit riskless bond
or swap, minus an option to exchange the bond or swap for the debtor's portfolio in bankruptcy. The
approach is on sound theoretical footing, practical for a firm with a trivial capital structure, but
questionable for a firm with a complex capital structure. The classic Black-Scholes (1973) paper and
Merton's paper on risky debt are the inspiration for this literature.

3. The reduced form approach multiplies the states of the world, beyond the usual ones with interest rate
derivatives, to include states where the debtor defaults. The risk neutral probability of default and the
recovery rate in default are important variables here. The approach is theoretically correct, but has a few
practical problems. The recovery rates, after default, are hard to come by. Deducing the risk neutral
probability of default depends on knowing the recovery rates, having a rich set of debt instruments for the
debtor, and assuming that default and the level of riskless rates is independent. Jarrow and Turnbull are
apparently the originators of this line of thinking. Many others have pursued this line of research.

Unfortunately, I would say that approaches 2 and 3 are "not ready for prime time". One is hard pressed to
come up with the required information on the debtor's portfolio to complete the structural model. One has
to make assumptions about the recovery rate and often about the risk neutral probability of default. I'm
not going to bet my money on either of these models at this time. However, one may not have the data
need to make approach 1 work, either. Making markets in credit derivatives is not for wimps or firms with
weak sales forces.

Complicated notation is a necessary evil in the world of credit derivatives, particularly for reduced form
models. Some authors seem impelled to use needlessly complicated notation, but sometimes the
complications are required. In either event, you just have to learn to love time and state subscripts. Best
wishes. – Dr. Risk

Market skepticism about agency credit quality? (11/28/98)

Question: My question has to do with the widening of the spread between 10 year treasuries and
mortgage rates. For as long as I can remember it has been 80/120 bases points over the 10 yr. T/B for a
30 year mortgage. A recent article in the NY Times said the spread has widened to over 200 B/P. If using
either Fannie or Freddie as the issuer of the MBS with the implied guarantee of the US government how
has the market assigned what appears to be a very high risk factor to this type of investment. [Today is
10/22/98.] I notice that the last question up-date was 9/28/98 so if you e-mail back a answer it would be
appreciated. – Paul

Answer: I won't address your facts. I'll assume they are correct and try to explain how that could be
consistent with rationality in the market. It's not clear that credit risk is the issue. I'll agree that Fannies
and Freddies have little of that, although they don't have federal government guarantees, as far as I
know, except for Fannie's limited "backstop authority".
The main issue here is the value of the borrower's prepayment option inherent in each loan that Fannie or
Freddie buys. The more valuable that option, the more the borrower will pay for it. The borrower pays
more by paying a larger coupon for a loan priced at par. The larger coupon leads to a larger yield to
maturity and spread over the ten-year Treasury. Apparently, right now, that option is extremely valuable,
which I don't find particularly amazing in this relatively uncertain world. See what has happened to
volatilities in the Treasury bond market. I haven't looked, but would guess that they have risen.

Maybe the following illustration will be useful. If you buy the ten-year, noncallable Treasury, trading at par,
then your coupon and yield to maturity will be a specific number, say 5% for ten years. (I'm making this
up. I don't know the real number and don't really care. If you don't like 5%, use another number and make
the appropriate changes, elsewhere.) If you had bought the ten-year, callable Treasury, trading at par,
and its coupon were also five percent, then three things could happen: 
(a) rates could rise and you would take an immediate capital loss, then receive your five percent, until
maturity
(b) rates could remain the same and you'd collect your coupon to maturity with no regrets
(c) rates could fall and the Treasury would exercise its call option and take your notes back, forcing you to
reinvest at a lower coupon. 
For the Treasury, it's a "heads, I win, and tails, the lender loses" situation. The ten-year, noncallable
Treasury with a coupon of 5% would be a poor investment for you, but a great funding coup for the the
Treasury. The Treasury would be happy to pay a higher coupon, in exchange for that call option, and
you'd be a fool not to demand it.

Similarly, the 30-year mortgage contains an embedded prepayment (American call) option. The lender
demands a high coupon to pay for selling that option, and the borrower is willing to pay for the option that
way. When the mortgage spread over the 10-year Treasury rises, it could be because of declining credit
quality or because of increased time value for the embedded call option. – Dr. Risk

Credit risk (7/28/98; revised 11/28/98)

Question: Dear  Dr. Risk – I research a litterature survey of the credit risk: How is credit risk measured?
– Stephane

Answer: Dear Stephane –  The short answer is, very badly. 

1. The classic approach is that of "fractional exposure", which is a forecast of potential market value of a
position – if positive. In its primitive form, this doesn't encompass correlation among credit risks. 
2. The state of the art model seems to be along the lines of VaR, but with more mathematics, combining
diffusion and Markov matrices. JP Morgan's CreditMetrics takes this approach. Currently, their model
assumes that the forward curve is constant, to isolate credit deterioration from market risk. However, most
credit deterioration comes from unfavorable market moves.
3. An actuarial approach, which starts by modeling default as a Poisson process, has its fans. CSFP's
CreditRisk+ takes this tack. I suggest that they rethink their model of correlation among defaults.
4. Default rates rise (fall) during a recession (boom). In principal, one might forecast the business cycle to
gain insight into default rates. Tom Wilson of McKinsey is pursuing this idea. Unfortunately, decades of
research have taught us only that the stock market predicts the business cycle as well as everything else
combined, so I think that Wilson's explorations will lead to dry holes. 
5. Default is ordinarily a rational decision, where equity holders decide not to exercise the option to keep a
firm alive. KMV has built an empire on this idea. The problem is quantifying the underlying asset's value
and volatility, as well as the appropriate "strike" price.
Just about everything I've seen is highly suspect  –  not the mathematical reasoning, only the assumptions
from which it derives. The problem is that much of the data you need to calibrate the models is not
observable. I hope to write more on the topic later this summer or early in the fall, and I'll be
offering courses on managing credit risk with credit derivatives in New York, Geneva, and London.
Meanwhile, look at the books on credit risk, mentioned above. – Dr. Risk