Sovereign default risk is a key risk for many other reasons than recent government rescues:

• Government bond markets represent some of the biggest counterparty exposures worldwide, particularly after the extensive deficit financing of the last two years
• Much of the financing of sovereign government debt is now being done by private sector lenders and other sovereigns, rather than multi-national development institutions
• Cross border financing is now the norm, not an exception, in today’s capital markets

For all of these reasons, more institutions, both private and public, have sovereign default risk exposure than ever before.

Until the launch of Kamakura’s sovereign default service in October 2008, ratings were the only real sovereign default risk assessment tool available.  Like corporate ratings, the ratings for sovereigns suffer from many problems when one is interested in very precise, quantitative risk assessment:

• Ratings lack an explicit maturity
• Ratings lack an explicit default probability
• Ratings change only infrequently, not in real time like bond spreads and credit default swaps
• Ratings are potentially subject to external political pressure, since governments have the power to regulate the raters
• Rating agency credit assessments of sovereigns are heavily weighted toward recent years and are almost exclusively focused on bonds, even though there are a wide range of borrowing tools used by sovereigns
• Ratings agencies are currently the victims of a crisis of confidence because of what many consider overly rosy ratings and conflicts of interest through the current credit crisis

For these reasons, an explicit default probability at key maturities is essential for the measurement of sovereign risk.

Best practice in sovereign default modeling is very similar to best practice for modeling corporate defaults:

• The default data base should be monthly to maximize the ability to measure the impact of macro factors on default risk
• Defaults should include defaults on all borrowing instruments, not just bonds
• There should be a full term structure of default probabilities derived from the modeling exercise
• There should be strict adherence to Basel II and corporate default definitions.  For many, being in arrears is “OK” if the borrower is a sovereign government.  Best practice calls 90 days past due a “fail” whether the borrower is a sovereign or not
• Macro-economic drivers of default should be extensively examined for statistical significance
• An extensive test regime to measure the accuracy of the prediction should be employed

The data base employed for the Kamakura Risk Information Services sovereign default service is a monthly data base which goes back to 1980, 10 years farther back into history than one can normally go with corporate default data bases.  The reason corporate default data bases are limited to 1990 onward is the lack of daily stock price data that commercial users are legally allowed to use for modeling.

There are substantial differences between sovereign default analysis and corporate default analysis:

• Much less sovereign debt is rated, compared to corporate debt, so there is much less data to work with in modeling
• “Cross default” clauses in lending to sovereigns have historically been much less common than in the corporate lending world.  In fact, sovereigns have much more in common with retail borrowers, who might be in default on a credit card but not on their mortgage. Rating agencies, for example, have often ignored defaults on sovereign bank debt as long as bond holders were being paid as scheduled.
• Much sovereign default has a “social welfare” purpose and this has in some cases affected whether or not an analyst would judge the borrowing to be in default

Because of these considerations, many studies of sovereign default have dramatically under-counted the number of default events or erred dramatically in identifying the date of default.  From an analytical point of view, there has historically been a strong feeling that one should treat sovereign defaulters kindly because analysts are rooting for the sovereign to succeed, hoping that the sovereign can lift its citizens from poverty.  Unfortunately, inaccurately assessing the risk of sovereign lending has the opposite effect in the long run.

The definition and timing of defaults in the Kamakura Risk Information Services sovereign default data base is the earliest of the following events:

• Being in arrears on interest by the World Bank definitions
• Being in arrears on principal by the World Bank definitions
• Being in arrears on borrowings from the International Monetary Fund by the IMF’s definition
• Declaration of the intent to default or actual default on any form of borrowing, regardless of the amount
• Being rated D or SD (selective default) by any major rating agency

A strict adherence to this definition of default results in much greater accuracy in the number and timing of defaults.  One has to be careful to adjust for “arrears” events that are clearly temporary, operational errors in funds transfers.  Both South Africa and China, for example, were in arrears by the World Bank definitions in recent years but these represented failed wire transfers or other issues, not true defaults.

For a data base beginning in 1980, one has to remember that a country that was in default prior to 1980 will have no entry in the data base until the original default was cured and the country becomes current on its borrowings, exactly as one would do for a corporate defaulter.  We also note that it is much for common for sovereigns than corporate for a given entity to be a “serial defaulter” so we need to track exits and entrances to the data base by the same country very carefully.

We look at case studies in tomorrow’s post.

Donald R. van Deventer

Kamakura Corporation

August 3, 2009