In what follows, for each example we plot rescheduled debt (the blue bar) and interest and principal in arrears (the red bar). The data graphed was provided by the World Bank. We apply the “failure” definition from our August 3 post, so the failure date is the earliest of an arrears event, a declaration of default, or a D or SD rating from a major rating agency.

The first example is Albania.

During 1991, Albania first fell into arrears, so 1991 would be the first occurrence of default and therefore it would be the “fail date” in a best practice sovereign default model that applies a modern Basel II “cross default” standard.  Since the arrears were never cured through 2004, Albania would be omitted from the data base from that moment onward, just as Enron was not included in the KRIS default data base after its failure in December 2001. Albania was not rated by S&P during this period.

Argentina is every economist’s favorite serial defaulter, with sovereign defaults dating back more than 100 years.  The graph below shows the key dates in the 1980-2004 time period:
 

Because Argentina was already in default at the start of the period, it does not appear as a non-defaulter until 1995.  While the scale of the graph makes it hard to see, Argentina went into arrears by the World Bank definition in 2000, but it was not rated SD by Standard & Poor’s until November 6, 2001.  Kamakura uses the earlier date as the default date.

Bangladesh has been continually in arrears during the 1980-2003 period.

It was not until 2004, its first year since 1980 with nothing in arrears, that Bangladesh would appear in the KRIS sovereign default data base.

Belize is another interesting case with three distinct failure dates:

S&P rated Belize an SD rating on December 7, 2006, the third of the three “fail dates.”

Botswana was in arrears from 1986 through the year end of 2000:

It’s a fairly common example, in contrast to corporations, of a failure event that is ultimately cured. After the “cure,” the company re-enters the data base as a non-defaulter.

The arrears data cannot be used uncritically, however.  The chart below for China is a case in point

China was reported as “in arrears” by the World Bank to the tune of only $300,000 at the year end 2003.  This was clearly just an operational mistake and not a true default.

Russia presents one of the most interesting cases for a sovereign default modeler, as one can see from the graph below.

The Russian Federation was created in 1991, and the country was already in arrears by the end of 1992.  Most analysts, however, focus on the bond defaults of 1998, and Standard & Poor’s did not issue an SD rating until January 27, 1999.  From a cross-default perspective, 1992 was the relevant date.

Another interesting case is that of Mexico.

The country has had two arrears “failure events,” but neither date coincides with the much publicized tequila crisis of 1994-1995.  Mexico was never rated D or SD during this period.

These examples show that the date of failure for sovereigns is strongly dependent on the definition of default.  Indeed, the data makes it clear that “sovereigns default because they want to,” tending to default on bank loans or loans from multinational lending organizations before they default on bonds.  In the same vein, the traditional view of retail borrowers has been that they would selectively fail to pay on credit cards while they continued to pay on their mortgage.

In a world where the cross-default clauses are rapidly being added to sovereign loan agreements, recognizing these earlier dates as “fails” is essential in predicting future sovereign defaults.  To ignore them would lower estimated default probabilities and cause a long delay in the prediction of a potential sovereign failure.

Donald R. van Deventer
Kamakura Corporation
Honolulu, August 4, 2009