Thursday, September 20, 2012

The Algorithm

"Does Standard & Poor take into account the impact of its ratings when assessing a country's credit worthiness?"

"No"
 - Myriam Fernandez de Heredia, S&P's director of sovereign ratings for Europe and Africa, on Los Desayunos de TVE (January 20th, 2012)

The new algorithm would change everything.  The Standard & Poor sovereign credit rating system used a hybrid-form of analysis that combined computerized quantitative modeling with expert qualitative insight.  Everyone understood, however, that the credit ratings themselves, once announced, changed people's behavior and, as a result, altered the equation.  It was a bright young man, a recent MIT graduate, that formulated an algorithm that incorporated the rating's projected impact into the final sovereign credit score.

For example, let us imagine that the credit rating for the sovereign bonds of the Republic of Krakhovia suffer a downgrade due to the country's high debt burden and low growth rate.  The country's S&P rating goes from A to BBB+.  The announcement spooks potential Krakhovian bond buyers; Krakhovia has trouble selling its bonds and must raise the bonds' interest rate; the rise in the interest rate exacerbates the country's already troubling debt levels.  So, the projected debt levels, once the impact of the BBB+ rating is taken into account, are higher than earlier models would have anticipated.  The new algorithm suggests Krakhovia's bond rating be downgraded to BBB.  A BBB rating, however, causes even greater concern than the original downgrade, sending bond interest rates even higher.  The new algorithm suggests Krakhovia's bond rating be downgraded to BB+.  BB+ bonds are borderline junk bonds, alarming bondholders who are increasingly trying to unload the assets.   The algorithm suggest a B- rating.  At B-, the bondholders are sent into a panic, causing a massive sell off and the crashing of the country's currency.  The algorithm suggest a CCC rating.  At CCC, the IMF enters the picture.

The algorithm, once tested, proved unusable because any downgrade (no matter how small) sent the rating spiraling to CCC (the lowest possible rating), and any upgrade resulted in an AAA rating (the highest possible).

The markets were never informed of the algorithm.

No comments:

Post a Comment