Confidence Game - Christine Richard [124]
In early 2008, Dinneen’s job was more tedious than entertaining. He was assigned to go through the model, cell by cell, making sure that the logic flowed and that the right cells connected. The process was tedious, partly because it took more than 40 minutes to rerun the model after making changes.
Dinneen would push the F9 button, walk away from the computer, return half an hour later, and find the model was still running. It required a leap of faith not to assume that the model had crashed his computer or become stuck in some infinite loop. “Excel thinks top down,” Dinneen says. “When it has to double back, when the connections are so circular, it takes a long time.”
The model ran through three iterations. The first time around, it peered into all of the CDOs and figured out what the losses on each one would be based on the projected losses on the approximately 100 securities referenced by a typical CDO. The program cycled through a second time, checking to see if any of the CDOs had exposure to other CDOs. When it found such CDOs, it docked the “outer” CDO for its share of any losses on the “inner” CDO. The final time through, the model reached down into the “inner” CDOs to see if they held exposure to CDOs. The “inner-inner” CDOs’ losses were passed up to the “inner” CDOs, and those losses were added to the total losses for the “outer” CDO. The three iterations gave a total forecast for losses on each directly guaranteed CDO. These losses were added to the direct guarantees on subprime mortgage- backed bonds, home-equity-loan deals, and Alt-A securities for a final tally.
The process gave Pershing Square amazing insight into the bond insurers’ portfolio, but it had to be accelerated. Paul Hilal, Ackman’s college friend who had joined the Pershing Square investment team in 2006, called his brother, Phil, who had recently bought a $16,000 customized personal computer for Christmas. Phil Hilal agreed to dash back to his apartment to run the Open Source Model on his computer.
CDOs such as Ridgeway Court Funding II Ltd, which was directly insured by Ambac, demonstrated the tangled web of exposures the model analyzed. Ridgeway Court had direct exposure to one of the worst-performing CDOs in the market, Carina CDO Limited. Ridgeway Court also had indirect interest in Carina through another CDO holding called 888 Tactical Fund Limited that was directly exposed to Carina. But it didn’t end there. Ridgeway Court also had exposure to two CDOs called Pinnacle Peak CDO Limited and Octonion CDO Limited that held interests in 888 Tactical Fund, which—if you are keeping track—had an interest in Carina.
While MBIA and Ambac claimed to have been highly selective in picking only the good CDOs to guarantee, there was so much overlap that they couldn’t escape the bad ones. There were 534 CDO deals backed by subprime mortgages originated between the beginning of 2005 and the end of 2007. MBIA backed just 25, and Ambac guaranteed 28. But the Open Source Model showed that MBIA was directly or indirectly exposed to 420 of the 534 CDOs, or 80 percent of the market, while Ambac had exposure to 389, or 73 percent of the market. When Phil Hilal pulled up the model and pressed the F9 button, it took just 30 seconds for his computer to run the program. The result was going to shock the market.
Backing securities on which there was supposedly no risk of default had appeared to be the perfect business for the bond insurers, Robert Fuller, who runs Capital Markets Management LLC, a Hopewell, New Jersey-based advisory firm, told me in December 2007. Now the business had “morphed into this monster that is devouring them.”
Chapter Twenty-One
Catastrophe and Revenge
It’s quite simple. The count is telling us horrible stories with the intention of making us all die of fear.
ALEXANDRE DUMAS,